HARDWARE

What runs on a regular computer without a GPU

Without a GPU an open model runs on the CPU and ordinary RAM. It is slower, but you buy nothing extra: a good fit where the answer is not needed instantly and volumes are modest — overnight document processing, transcription, knowledge-base search.

500 familiesUpdated 22 Sep 2026Calculate video memory for your model size

What this means in practice

Office laptop16 GB of RAM, a regular CPU. Handles small language models in compressed form, speech recognition, speech synthesis and search embeddings.
Workstation or mini PC32-64 GB of RAM. The same, but a larger model and several tasks at once.
Virtual server without a GPUThe cheapest hosting option. Good for background work: process the mailbox overnight, index an archive, transcribe call recordings.

What runs on it

Text 74

TextGGUF2024–2026

Sarvam

Sarvam AI · India

Indian models focused on 22 languages of India. Sarvam 30B and 105B (2026) are MoE models with strong reasoning and agent skills. In November Sarvam-Prism came out (Apache 2.0, based on Sarvam-105B): a decision model that, given a context and questions with fixed answer options, returns the probability of each in one pass and writes no text. It needs a graphics card with FP8 and about 120 GB of memory (B200 or H200); a paid API exists too.

  • Multilingual customer support
  • Reasoning and calculation tasks
  • Agents with tool calling
Sizes
2B – 105B-A10B
Hardware
from: Laptop
Commercial use allowedDetails
TextOllama2024–2026

NVIDIA Nemotron

NVIDIA · USA

NVIDIA models for agents and reasoning, optimized to run fast on its GPUs. Nemotron 3 is a Mamba and MoE hybrid from 4B to 550B; Nano Omni handles video, audio and images (English only). The new Nemotron 3.5 Super VL (120B, 12B active, up to 1M-token context) reads text, images and video up to two minutes, with a focus on documents and agents; the compressed Bolt (74B, 9B active) fits into 24 GB of video memory. Super VL languages: English, Spanish, French, German, Italian, Japanese. Dory 1.0 (4B, up to 256K-token context, English) runs the same block of layers several times, and the number of passes can be changed per task: more passes give more accuracy without a new model and without long reasoning in the text.

  • Agents with tool calling
  • Reasoning and calculation tasks
  • Answers based on long documents
Sizes
4B – 550B-A55B
Hardware
from: Laptop
Commercial use with conditionsDetails
Text2025–2026

Ling / Ring

Ant Group (inclusionAI) · China

An Ant Group family: Ling for standard models, Ring for reasoning ones. There are trillion-parameter flagships and the efficient Ling-3.0-tiny, which needs only 1.3 billion active parameters. The new Ling-3.1-flash (560B, 25B active, 1M-token context) targets agents, office tasks and development. The improved Ling-3.1-flash-1019 (MIT) raised the Artificial Analysis index from 41 to 44 by the authors estimate.

  • Corporate assistant
  • Agents for office processes
  • Financial analytics (Fin version available)
Sizes
7.9B-A1.3B – 1T
Hardware
from: Laptop
Commercial use allowedDetails
TextRUOllama2025–2026

Liquid LFM

Liquid AI · USA

Models with a new architecture for on-device use: fast on a regular CPU and on phones. Versions for data extraction, RAG and tools, plus LFM2.5-VL for images and voice LFM2.5-Audio. The largest, LFM2.5-24B-A2B (2.45B active, 131K-token context, Russian in the language list), always reasons before answering and targets fast on-device agents. In October the LFM3 generation came out on a new tri-hybrid architecture: LFM3-1.7B for local agents (256 thousand tokens of context, about 1 GB of memory at 4 bits, pre-trained on 22 trillion tokens), LFM3-VL-2B for images, LFM3-Audio-2B for speech and LFM3-Omni-2B, which understands text, images, video and sound and answers in text and speech. The authors do not recommend it for complex code and multi-step reasoning.

  • Offline assistant on a laptop or phone
  • Data extraction from documents
  • Tool calling in apps
Sizes
230M – 24B-A2B
Hardware
from: Laptop
Commercial use with conditionsDetails
Text2026

Spark-X

XHToken (the Spark project) · not stated

A compact 4B model to run right on a device, focused on agents: tool calling, multi-step planning, programming, long context. Version 2.6 came about a month after 2.5 with the same architecture and improved only through further training. The reasoning budget is adjustable: none, low, high. The authors state that it holds its own against models of around 9B. Languages on the card: English and Chinese, Russian is not stated. Running it needs the model's own code in transformers. The figures come from the authors. At the end of October a 1.7B size was added (INT4, INT8, FP8).

  • an agent with tool calling on a home computer or laptop
  • a local programmer assistant that does not need a big server
  • AI built into an app with a choice of how much the model thinks
Sizes
1.7B и 4B
Hardware
from: Laptop
Commercial use allowedDetails
Text2026

Humanizer

jialinyyzz (sgaofen) · not stated

A 12B model that rewrites drafts written by AI so that they read like a person wrote them: emails, essays, reports, forum posts. It is trained to keep every number, unit, date, name and quote and to add nothing. It runs on your own computer; there is a ready app for Mac and Windows that works offline, and GGUF files of different compression for llama.cpp. Stated languages: English and Chinese, Russian is not stated. By the author's measurements, the Originality.ai detector judged 95% of the English rewrites human. The result still has to be read through, especially numbers and names.

  • rewriting drafts of emails and reports so that they sound more natural
  • editing texts after AI without losing numbers, dates and quotes
  • local processing of texts that must not be sent to the cloud
Sizes
12B
Hardware
from: Laptop
Commercial use allowedDetails
TextOllama2026

Underdog Saluki

ConwayResearch (Underdog) · not stated

Qwen3.8-27B squeezed to 2 bits into a single 7.9 GB GGUF file instead of 54 GB, with the focus on keeping tool calling intact. By the authors' measurements (their Underdog Bench), tool calling is 88 against 84 for the full model, parallel calls 42 against 35, and on average 96% of the full model's quality across nine tests. It runs on plain llama.cpp and Ollama, thinking mode can be switched on and off, and it understands images with a separate vision file (about 0.6 to 0.9 GB). Language stated on the card: English.

  • an agent with tool calling on a home graphics card
  • a local coding assistant with functions
  • analysing pictures with the vision file attached
Sizes
27B (7.9 GB file)
Hardware
from: Laptop
Commercial use allowedDetails
TextOllama2023–2026

DeepSeek

DeepSeek · China

DeepSeek's flagship line: from the first 7B/67B to V4-Pro with 1.6 trillion parameters. Closed-model quality under an open MIT license; V4-Flash-Vision-Exp and V4.1-Flash understand images, context up to 1M tokens. The fresh V4.1-Flash-1017 answers shorter and faster than the previous Flash: by the measurements of DeepSeek itself, 77.0 vs 74.2 on DeepSWE v1.1 and 42.3 vs 31.2 on Terminal-Bench 4.0.

  • Employee assistant on your own server
  • Analysis of long contracts and reports
  • Agents that work with tools and APIs
Sizes
7B – 1.6T-A49B
Hardware
from: Laptop
Commercial use allowedDetails
Text2026

AstaBrief

Ai2 (Allen Institute) · USA

An 8B model that turns a question and the excerpts of scientific papers you give it into a report with citations. It does not search by itself: excerpts must be supplied with the question.

  • literature overview with source citations
  • summary of your own set of PDFs
  • draft analytical note with quotes
Sizes
8B
Hardware
from: Laptop
Commercial use allowedDetails
TextRU2025–2026

Лаборатория ИИ СВФУ: якутский язык

AI Laboratory of the North-Eastern Federal University · Russia

The most complete set of models for the Sakha (Yakut) language in one place: language models of several sizes, translation to and from Russian, embeddings and text recognition from images. Quality varies, so test it on your own data.

  • Answering and writing in the Sakha language
  • Translating documents and materials
  • Semantic search over Sakha texts
Sizes
1.1B – 27B
Hardware
from: Laptop
Commercial use with conditionsDetails
TextRUOllama2024–2026

Aya

Cohere Labs · Canada

Multilingual models from Cohere's research arm, covering 23 to 100+ languages. Tiny Aya (2026, 3.3B) runs on a regular PC, but for non-commercial use only.

  • Translation and correspondence in less common languages
  • Multilingual chat assistant
  • Analysis of images with text (Vision)
Sizes
3.3B – 35B
Hardware
from: Laptop
Commercial use with conditionsDetails
TextRU2024–2026

GigaChat

Sber · Russia

Sber open models with strong Russian language support and local context, from 10B-A1.8B to 702B, all MIT. GigaChat3.1-Audio handles recordings up to two hours; GFusion is a fast diffusion text version.

  • Russian-language employee assistant on your own server
  • Customer replies and request handling in Russian
  • Working with contracts and internal policies
Sizes
10B-A1.8B – 702B-A36B
Hardware
from: Laptop
Commercial use allowedDetails

Text analysis 27

Text analysisRU2026

Liquid d1

Liquid AI · USA

Small decision models: from text, an image or a voice clip they answer yes or no, pick an option or give a rating in one pass, without writing text. d1-3B sees images, d1-omni-600M also hears speech up to 30 seconds. Russian is listed. The open d1-2B (text, pictures and video, based on LFM3-VL-2B) and d1-Omni-2B (sound as well) decide in milliseconds, 5.7 ms on an H100 by the authors measurements; besides choosing an answer they pull out text, numbers and dates with their positions and find objects in a picture. It is not a chat model.

  • routing requests and tickets to teams
  • checking a photo for defects or violations against a list of questions
  • routing voice commands
Sizes
600M, 2B and 3B
Hardware
from: Laptop
Commercial use with conditionsDetails
Text analysisRU2026

vLLM Semantic Router: Decision и Vela

vLLM Semantic Router project · international project

An open family of models for an AI gateway: they answer a request with probabilities without generating text. They detect topic and urgency, jailbreak attempts, harmful requests, personal data and whether a tool is needed. Small versions run on a CPU, Vela understands 17 languages including Russian. The new d3 (27B) also reads several images and videos, and d3-turbo is, by the authors' own estimate, 4 points more accurate on text. The newest in the line, d3-ultra, came out on October 24.

  • routing requests between models and teams
  • checking a request for a jailbreak attempt or harmful content before the main model
  • finding personal data with positions in a text
Sizes
0.3B – 27B
Hardware
from: Laptop
Commercial use allowedDetails
Text analysis2026

LIDstral Arabic

Mistral AI · France

Quickly identifies the language and dialect of text in Arabic script: Modern Standard Arabic, all dialects and 51 classes overall. It tells apart close languages in the same script (Persian, Urdu, Pashto and others) so that a text is not called Arabic just because of its letters. Runs on a CPU.

  • detecting the dialect in customer requests in Arabic
  • sorting texts by language in Arabic script
  • preparing data for training and search
Sizes
a fastText model with calibration, parameter count not stated
Hardware
from: Laptop
Commercial use allowedDetails
Text analysis2026

Intern-Decision

Shanghai Artificial Intelligence Laboratory · China

A decision model based on Qwen3.5-4B: from text and images it answers all questions about a situation with probabilities in one pass, without generating text. Suits routing requests and rating urgency.

  • routing requests to teams
  • rating priority and urgency of requests
  • checking from a photo that everything is in order
Sizes
4B
Hardware
from: Laptop
Commercial use allowedDetails
Text analysis2026

Julia

Supersonic Labs · not stated

A very small decision model: from a text, a question and answer options it picks one option, gives a rating or answers yes or no. Runs on an ordinary CPU and in the browser.

  • routing requests by topic
  • detecting the tone of reviews
  • fast classification of requests
Sizes
144M
Hardware
from: Laptop
Commercial use allowedDetails
Text analysisOllama2026

Clef

Cloudflare · USA

Decision models: they do not write text but read a situation (text, JSON, an image, video frames, and for Clef-Omni also audio) and return in one pass the probability of each answer to typed questions: yes or no, pick one, rate on a scale. The agent gets ready numbers instead of parsing free text.

  • sorting requests and tickets into categories with a probability
  • checking a document or photo against a list of questions (is everything stated, is there a violation)
  • choosing the next step of an agent without parsing a text answer
Sizes
27B, 9B (Clef-flash) and 30B-A3B (Clef-Omni)
Hardware
from: Laptop
Commercial use allowedDetails
Text analysisRU2026

BashkirRoBERTa и набор моделей для башкирского

Tvoy_Tezka (failed09) · not disclosed

A small toolkit for Bashkir text: a fill-mask language model, a Bashkir language detector, fastText embeddings, a Bashkir-Russian pair scorer and a letter restorer. All published in autumn 2026, so check quality on your own data.

  • Spellchecking and word suggestion in Bashkir
  • Separating Bashkir texts from other languages
  • Restoring language-specific letters in text
Sizes
50M for BashkirRoBERTa
Hardware
from: Laptop
Commercial use with conditionsDetails
Text analysis2026

FlowX OpenNER

FlowX.AI · USA and Romania

A family of small models that pull amounts, deal parties, jurisdictions, accounts and sanctions-screening mentions out of banking and insurance texts. A human makes the decision on a disputed operation; automatic blocking without review is not acceptable.

  • Extracting amounts, parties and jurisdictions from operation descriptions
  • Parsing customer due diligence material and sanctions lists
  • Marking account numbers and payment details in documents
Sizes
about 150M (based on ModernBERT-base)
Hardware
from: Laptop
Commercial use allowedDetails
Text analysisGGUF2026

FlowX ScamGuard

FlowX.AI · USA and Romania

Small models from a banking software company that read the text of an SMS, email or chat message and explain which scam tactic it uses. A human makes the decision on a disputed operation; automatic blocking without review is not acceptable.

  • Flagging incoming messages that show signs of deception
  • Explaining which phrase triggered the flag
  • Filtering emails and messages before they reach an operator
Sizes
0.6B – 1.7B
Hardware
from: Laptop
Commercial use allowedDetails
Text analysis2024–2026

GLiNER

Urchade Zaratiana and Fastino AI · France / USA

Finds the entities you need in text without training: just list what to look for (name, amount, date). GLiNER2 also classifies text. Multilingual versions understand Russian.

  • Extracting names, amounts and dates from emails and contracts
  • Parsing requests into CRM fields
  • Classifying requests by topic
Sizes
about 50M to 500M
Hardware
from: Laptop
Commercial use allowedDetails
Text analysisOllama2024–2026

NuExtract

NuMind · France

Models for template-based data extraction: give it a document or scan and a JSON field template, get a filled-in JSON back. NuExtract3 (4B) also converts scans to Markdown.

  • Extracting company details, amounts and dates from invoices and contracts into JSON
  • Parsing receipts, waybills and forms against a set template
  • Converting scans to Markdown for search
Sizes
0.5B – 8B
Hardware
from: Laptop
Commercial use allowedDetails
Text analysisRU2026

Tatar NLP Community

Tatar NLP Community · not disclosed

A recent set of models for Tatar: morphological analysis on several base models, question answering about Tatar and Russian place names, and Mistral and GPT-2 fine-tunes for Tatar.

  • Morphological analysis of Tatar texts
  • Answering questions about place names
  • Generating text in Tatar
Sizes
178M for the morphology models, 7B for the Mistral fine-tune
Hardware
from: Laptop
Commercial use with conditionsDetails

Text to speech 26

Text to speech2026

Ming-3.0-tts

inclusionAI (Ant Group) · China

A family of open speech models from Ant Group: a dense 0.5B based on Qwen2.5-0.5B and the MoE version tiny (7.9B parameters, about 1.3B active, based on Ling V3). They clone a voice from a short sample, create a voice from a description in words, and control delivery, emotion, speed, pitch and volume. There are more than 100 built-in Chinese voices, five English ones and 14 inline tags for a breath, cough, laugh, sigh and hesitation. Languages: Chinese, English, German, Spanish, French, Korean, Japanese and the Sichuanese and Cantonese dialects, no Russian. Sample rate is 44.1 kHz. Running it needs the code from the GitHub repository, and the Hugging Face page only distributes the weights. There is a Comfy-Org build for ComfyUI.

  • voicing videos and podcasts in Chinese and English with emotion control
  • cloning a voice from a short sample
  • creating a voice from a description for games and apps
Sizes
0.5B and 7.9B-A1.3B (tiny)
Hardware
from: Laptop
Commercial use allowedDetails
Text to speechRU2024–2026

Fish Speech / OpenAudio

Fish Audio · USA / China

Speech synthesis with voice cloning and emotion control in 80+ languages, including Russian. Quality is close to paid services, but the weights are for research only. The new Drama 3 (35B, 3B active) voices scenes with several characters (up to 16 voices in one text), understands tags such as a sigh or a laugh, over 80 languages, Russian is in the list.

  • Voice cloning
  • Emotional voiceover
  • Multilingual voiceover
Sizes
0.5B – about 4.5B
Hardware
from: Laptop
Non-commercial onlyDetails
Text to speechGGUF2025–2026

IndexTTS

bilibili · China

Speech synthesis with voice cloning and precise duration control, handy for video dubbing. Controls emotion separately from timbre.

  • Video dubbing matched to timing
  • Voice cloning
  • Emotional voiceover
Sizes
about 1B – 2B
Hardware
from: Laptop
Commercial use with conditionsDetails
Text to speechRUGGUF2025–2026

Zonos

Zyphra · USA

Speech synthesis with voice cloning and fine control over emotion, speed and pitch.

  • Voice cloning
  • Emotional voiceover
  • Voicing videos
Sizes
about 1.6B
Hardware
from: Laptop
Commercial use allowedDetails
Text to speechRUGGUF2025–2026

Chatterbox

Resemble AI · USA

Speech synthesis with voice cloning and adjustable expressiveness. The multilingual version supports 23 languages, including Russian; Turbo and Flash are sped up for live dialogue.

  • Voice for a bot or assistant
  • Cloning a brand voice
  • Voicing videos
Sizes
about 350M – 500M
Hardware
from: Laptop
Commercial use allowedDetails
Text to speechRU2025–2026

MOSS-TTS / MOSS-TTSD

OpenMOSS (Fudan University) · China

A speech synthesis family: multi-voice dialogue voicing (TTSD), fast synthesis for live conversation and the tiny Nano. Version 1.5 supports 30+ languages, including Russian.

  • Voicing podcasts and dialogues
  • Voice for an assistant
  • Voice cloning
Sizes
100M – 8.5B
Hardware
from: Laptop
Commercial use allowedDetails
Text to speechRU2025–2026

Supertonic

Supertone · South Korea

Very fast, lightweight speech synthesis that runs directly on the device, without a GPU or the cloud. Supertonic 3 speaks 31 languages, including Russian.

  • Voicing voice bot replies on an ordinary server
  • Voiceover in offline and mobile apps
  • Reading texts and notifications aloud
Sizes
about 99M
Hardware
from: Laptop
Commercial use with conditionsDetails
Text to speech2025–2026

Kyutai STT, TTS и Pocket TTS

Kyutai · France

Streaming speech recognition and synthesis models from the makers of Moshi: they start speaking and transcribing without waiting for the end of a phrase. Pocket TTS (100M) runs on a CPU. English, French and a few other European languages, no Russian.

  • Streaming speech transcription for voice bots
  • Voicing replies with minimal delay
  • Speech synthesis on a server without a GPU (Pocket TTS)
Sizes
100M (Pocket TTS) – 2.6B
Hardware
from: Laptop
Commercial use allowedDetails
Text to speechRUGGUF2025–2026

VoxCPM

OpenBMB (ModelBest, Tsinghua University) · China

Speech synthesis with voice cloning and natural intonation. VoxCPM2 supports 30 languages, including Russian.

  • Voice cloning
  • Voicing videos and audiobooks
  • Voice for an assistant
Sizes
0.5B – 2.3B
Hardware
from: Laptop
Commercial use allowedDetails
Text to speechRU2026

OmniVoice

k2-fsa (Next-gen Kaldi) · China

Speech synthesis with voice cloning from a short sample in 646 languages, including Russian and languages of Russia's peoples. A voice can be described in words. Weights are for non-commercial use only.

  • Voiceover in rare languages
  • Voice cloning from a sample
  • Research and prototypes of multilingual voiceover
Sizes
0.6B
Hardware
from: Laptop
Non-commercial onlyDetails
Text to speechRUGGUF2026

Qwen3-TTS

Alibaba (Qwen) · China

Speech synthesis in 10 languages, including Russian: voice cloning from 3 seconds, ready-made voices and creating a voice from a text description.

  • Voice for a bot or assistant
  • Cloning a brand voice
  • Choosing a voice by description
Sizes
0.6B – 1.7B
Hardware
from: Laptop
Commercial use allowedDetails
Text to speechRU2024–2025

CosyVoice / Fun-CosyVoice

Alibaba (Tongyi, FunAudioLLM) · China

Speech synthesis with voice cloning from a short sample and streaming output for live dialogue. Version 3 supports 9 languages, including Russian.

  • Voice for a bot or assistant
  • Cloning a brand voice
  • Voicing videos
Sizes
300M – 0.5B
Hardware
from: Laptop
Commercial use allowedDetails

Documents and OCR 25

Documents and OCROllama2026

MiniCPM-OCR

OpenBMB (ModelBest) · China

The first MiniCPM model built only for documents: it reads text, tables, formulas, charts and seals, and pulls the needed fields out of papers. Just 1.3B. By the authors' tests, 93.68 on Real5-OmniDocBench (document parsing in real conditions: scans, photos, skewed and recaptured pages) and 54.7 on OCRBench v2. Chinese and English are stated, Russian is not. An auxiliary DSpark model is published for it that speeds up generation: according to the authors processing is 55% faster, and the speed is about 1.5 times that of PaddleOCR-VL-1.6 and 3.6 times that of MinerU2.5 Pro. It runs in vLLM, SGLang, llama.cpp and Ollama, with compressed GGUF, AWQ and GPTQ versions.

  • parsing scans and photos of documents into markup: text, tables, formulas
  • extracting fields from invoices, contracts and forms in Chinese and English
  • fast processing of a document flow on a small server
Sizes
1.3B
Hardware
from: Laptop
Commercial use allowedDetails
Documents and OCR2026

Unlimited-OCR

Baidu · China

Baidu's OCR model building on DeepSeek-OCR ideas: processes multi-page documents and PDFs in a single pass and outputs structured text. Claimed to be multilingual, but the language list is not published. On October 27 Unlimited-OCR-2 (MIT, about 3.3B) came out: according to the authors it parses documents of unlimited length (the SR-SWA technique), and the MarathonDocBench set for very long documents was published with it.

  • Converting multi-page PDFs and scans to text and Markdown
  • Recognizing contracts, invoices and reports
  • Preparing document archives for search and RAG
Sizes
3.3B
Hardware
from: Laptop
Commercial use allowedDetails
Documents and OCRRU2026

Indra OCR

Datalab · USA

A tiny OCR model that runs on an ordinary CPU with no graphics card and no PyTorch: it uses ONNX Runtime, takes about 230 ms per page on four cores and reads up to 49 pages per second. It turns PDFs and images into markdown, plain text or JSON with boxes for lines and blocks (headings, lists, tables, captions) and reading order. In digital PDFs it takes the existing text layer and recognises only what is missing. Russian is among the stated languages. For handwriting, formulas and complex tables the authors point to their larger Chandra. The speed and accuracy figures are the authors' own.

  • fast bulk conversion of PDFs and scans to markdown on a server with no graphics card
  • search over document archives: text with boxes and reading order
  • pre-marking pages before a heavy model
Sizes
about 1M parameters (the models weigh about 18 MB)
Hardware
from: Laptop
Commercial use with conditionsDetails
Documents and OCR2026

WeVisDoc

Tencent · China

Parses a whole document page: turns it into Markdown with LaTeX formulas and HTML tables. The 4B version scores 95.38 on OmniDocBench v1.6. Languages on the card: English and Chinese.

  • turning scans and PDFs into Markdown for search and RAG
  • recognizing tables and formulas
  • parsing reports and scientific papers
Sizes
2B и 4B
Hardware
from: Laptop
Commercial use allowedDetails
Documents and OCRGGUF2025–2026

LightOnOCR

LightOn · France

A French OCR model that converts a page into text in one pass and is fast on high volumes. The third generation (0.8B, 1B, 4B) also returns block coordinates, image descriptions and chart data as a table. Languages on the version 2 card: European languages, Chinese and Japanese; no Russian.

  • Recognising invoices and contracts in European languages
  • Table recognition
  • Converting PDFs to text for search and RAG
Sizes
0.8B – 4B
Hardware
from: Laptop
Commercial use allowedDetails
Documents and OCR2026

TeleOCR

TeleAI (China Telecom) · China

A new lightweight document parsing model that led the OmniDocBench v1.6 benchmark at release. Handles pages photographed on a phone and crumpled pages well. Languages on the card: Chinese, English, Japanese.

  • Recognising invoices and delivery notes photographed on a phone
  • Recognising tables and formulas
  • Converting documents to Markdown for RAG
Sizes
about 1.2B
Hardware
from: Laptop
Commercial use allowedDetails
Documents and OCRRUGGUF2025–2026

HunyuanOCR

Tencent · China

A lightweight OCR model from Tencent: document parsing, finding text in photos, field extraction and translating text from images. Version 1.5 is faster and runs on an ordinary PC.

  • Extracting fields from invoices and delivery notes
  • Recognising tables and formulas
  • Translating text in photos and scans
Sizes
1B
Hardware
from: Laptop
Commercial use with conditionsDetails
Documents and OCRRU2024–2026

Surya OCR

Datalab · USA

A compact OCR toolkit from the makers of Marker and Chandra: text recognition, page layout, reading order and tables. Surya OCR 2 (650M) also runs on a CPU, and Russian is among the supported languages.

  • Recognizing scans and PDFs, including in Russian
  • Page layout: headings, tables, images, reading order
  • Recognizing tables by rows and columns
Sizes
up to 650M
Hardware
from: Laptop
Commercial use with conditionsDetails
Documents and OCRRU2022–2026

PP-OCR и PP-DocLayout (классический PaddleOCR)

Baidu (PaddlePaddle) · China

Classic lightweight PaddleOCR models: detecting and recognizing lines of text plus page layout. They run on CPUs and phones; there is a separate model for East Slavic languages, including Russian.

  • Recognizing text on scans, photos and screens
  • Reading labels, displays and markings in production and warehouses
  • Page layout: tables, formulas, stamps, headings
Sizes
from 1.5M to tens of millions of parameters
Hardware
from: Laptop
Commercial use allowedDetails
Documents and OCRGGUF2025–2026

MinerU

Shanghai AI Laboratory (OpenDataLab) · China

A popular open tool for converting PDFs to Markdown with its own small model. MinerU2.5-Pro was improved through data alone, without growing in size. Languages on the card: Chinese and English.

  • Converting PDF reports and contracts to Markdown
  • Recognising tables and formulas
  • Preparing documents for RAG and search
Sizes
0.9B – 1.2B
Hardware
from: Laptop
Commercial use allowedDetails
Documents and OCRRUGGUF2025–2026

PaddleOCR-VL

Baidu (PaddlePaddle) · China

A compact document parsing model from the popular PaddleOCR toolkit. Per the model card it supports 109 languages, including Russian; version 1.6 leads the OmniDocBench benchmark.

  • Recognising invoices, contracts and delivery notes, including in Russian
  • Recognising tables, formulas and stamps
  • Converting scans to Markdown and JSON
Sizes
0.9B
Hardware
from: Laptop
Commercial use allowedDetails
Documents and OCRRUGGUF2025–2026

Chandra OCR

Datalab · USA

A strong OCR model from the authors of Marker and Surya: handwriting, forms, tables. Per the model card it supports 90+ languages, with Russian among the examples.

  • Recognising invoices, contracts and delivery notes, including in Russian
  • Recognising handwritten forms and questionnaires
  • Recognising complex tables
Sizes
5B – 9B
Hardware
from: Laptop
Commercial use with conditionsDetails

Search and RAG 24

Search and RAGRU2025–2026

NVIDIA Nemotron Embed

NVIDIA · USA

NVIDIA embeddings for search and RAG. Nemotron-3-Embed, released in 2026, is under the permissive OpenMDW license and works in many languages. In October Nemotron-3.5-Embed (1B and 8B, OpenMDW 1.1) came out: it finds documents by text and by the image of a page with tables, charts and infographics; for text it was tested in 34 languages, Russian among them. The 8B is built on Ministral 3.

  • Search across corporate documents
  • RAG for chatbots and assistants
  • Search across images and pages (VL versions)
Sizes
1B – 8B
Hardware
from: Laptop
Commercial use with conditionsDetails
Search and RAG2026

Trident

OpenBMB · China

Embeddings for mixed-modality search: text, images and pages with text and a picture together land in one space, and a single text query can rank them in one shared list. Built on research showing that ordinary models prefer text over images. Language on the card: English.

  • searching a catalog where some items are described by text and some by a picture
  • searching slide decks and pages with diagrams
  • RAG over a mixed knowledge base
Sizes
2B (based on Qwen3-VL-2B)
Hardware
from: Laptop
Commercial use allowedDetails
Search and RAG2026

WeMM-Embedding

Tencent · China

Universal embeddings based on Qwen3.5: they turn text, images, video and visual documents into one vector of 4096 numbers. Audio is not supported. Languages on the card: Chinese and English.

  • searching a product catalog by picture or description
  • searching video and slide decks
  • RAG over mixed text and image materials
Sizes
2B и 9B
Hardware
from: Laptop
Commercial use allowedDetails
Search and RAGOllama2025–2026

EmbeddingGemma

Google · USA

Small multilingual embeddings from Google that run on a phone or laptop without internet. The second version also understands images, video and audio in one vector space.

  • On-device document search
  • RAG without sending data outside
  • Search over photos, video and audio by a text query
Sizes
300M и 740M (EmbeddingGemma 2)
Hardware
from: Laptop
Commercial use with conditionsDetails
Search and RAGGGUF2026

pplx-embed

Perplexity · USA

Embeddings from the Perplexity search service. Some versions take into account the context of the whole document, not just a single fragment.

  • Search across large document collections
  • RAG that accounts for document context
  • Website and catalog search
Sizes
0.6B – 9B
Hardware
from: Laptop
Commercial use allowedDetails
Search and RAGRU2024–2026

FRIDA / Giga-Embeddings

Sber (SberDevices) · Russia

Sber embeddings built for Russian: according to the developers, among the best on Russian-language search benchmarks. FRIDA is compact, Giga-Embeddings is more powerful.

  • Search across Russian-language documents
  • RAG for chatbots in Russian
  • Classifying requests and reviews
Sizes
480M – 10B-A1.8B
Hardware
from: Laptop
Commercial use allowedDetails
Search and RAGRUGGUF2023–2026

Jina Embeddings

Jina AI · Germany

Strong multilingual embeddings with long context; v5-omni understands text, images and audio. Recent versions are open for non-commercial use only.

  • Search across documents in many languages
  • Search across images and scans
  • Classification and clustering
Sizes
33M – 3.8B
Hardware
from: Laptop
Commercial use with conditionsDetails
Search and RAGRUOllama2024–2026

Granite Embedding

IBM · USA

Lightweight IBM embeddings for enterprise search, trained on data with clear rights. R2, released in 2026, became multilingual.

  • Search across corporate documents
  • RAG on a regular server without a GPU
  • Reranking results
Sizes
30M – 311M
Hardware
from: Laptop
Commercial use allowedDetails
Search and RAGGGUF2025–2026

Octen Embedding

Octen · USA / Singapore

Qwen3-Embedding models fine-tuned by the startup Octen for search in legal, financial and medical texts. As of January 2026 the 8B version topped the RTEB leaderboard.

  • Search across contracts and case law
  • Search across financial reports
  • Search across long documents up to 32K tokens
Sizes
0.6B – 8B
Hardware
from: Laptop
Commercial use allowedDetails
Search and RAGRU2026

Harrier (harrier-oss)

Microsoft · USA

Microsoft's 2026 multilingual embeddings with context up to 32K tokens; Russian is on the language list. The 270M and 0.6B versions run on a regular server, 27B is the most accurate.

  • Multilingual knowledge base search
  • Picking passages for RAG
  • Search across long documents
Sizes
270M – 27B
Hardware
from: Laptop
Commercial use allowedDetails
Search and RAGRUOllama2025–2026

Qwen3 Embedding / Reranker

Alibaba (Qwen) · China

Embeddings and rerankers based on Qwen3, among the best open ones for multilingual search, including Russian. VL versions search images, screenshots and video.

  • Knowledge base search for RAG
  • Reranking results before answering
  • Search across scans, slides and screenshots
Sizes
0.6B – 8B
Hardware
from: Laptop
Commercial use allowedDetails
Search and RAG2026

Voyage 4 nano

Voyage AI (MongoDB) · USA

The only open model in the Voyage 4 line: its vectors are compatible with the paid larger versions, so you can start locally and move to the API later.

  • Document search on your own server
  • RAG for small knowledge bases
  • Finding similar texts
Sizes
about 340M
Hardware
from: Laptop
Commercial use allowedDetails

Speech to text 18

Speech to textRU2023–2026

NVIDIA Parakeet / Canary / Nemotron Speech

NVIDIA · USA

Fast NVIDIA speech recognition models, including streaming ones for real-time use. Parakeet TDT v3 and Nemotron 3.5 ASR understand Russian. The new Nemotron 3.5 Transcribe (34B, 4.3B active, OpenMDW 1.1) transcribes recordings of up to an hour in 40 languages, translates speech, splits it by speaker, puts word timestamps and answers questions about the content (questions in English only); all tasks are available for Russian. Ready files only, live streaming is not supported. The figures are the authors: 5.33% word errors on English in the open Hugging Face leaderboard.

  • Transcribing calls and meetings
  • Video subtitles
  • Real-time voice input
Sizes
110M – 34B-A4.3B
Hardware
from: Laptop
Commercial use with conditionsDetails
Speech to text2026

Cactus Whistle

Cactus Compute · not stated

Speech recognition in a single 16.9 MB file: runs on a phone, watch, robot and CPU without a graphics card. Languages: English, German, French, Spanish, Italian, Dutch, Polish. Gives word timestamps.

  • voice input in mobile apps without internet
  • controlling a device by voice
  • highlighting words in subtitles by timestamps
Sizes
16.9 MB
Hardware
from: Laptop
Commercial use allowedDetails
Speech to textRUGGUF2025–2026

Voxtral

Mistral AI · France

Mistral's speech models: they understand audio, transcribe and answer questions about a recording. The Realtime version recognizes speech live and supports Russian; speech synthesis and a separate version for Arabic dialects are also available.

  • Transcribing and summarizing recordings
  • Asking questions about audio
  • Real-time recognition
Sizes
3B – 24B
Hardware
from: Laptop
Commercial use with conditionsDetails
Speech to text2026

Phonon

Fermion Research · not stated

A very compact English speech recognition model: 164 MB, close in accuracy to its large Parakeet base. Transcribes an hour of audio in about 20 seconds on a laptop.

  • transcribing English meetings and calls
  • dictating text on a laptop offline
  • subtitles for English videos
Sizes
164 MB (about 2.1 bits per weight)
Hardware
from: Laptop
Commercial use allowedDetails
Speech to textRUGGUF2025–2026

VibeVoice

Microsoft · USA

Microsoft speech models: long multi-voice dialogue synthesis, fast synthesis for live conversation, and recognition of long recordings split by speaker, including in Russian.

  • Transcribing long meetings with speaker labels
  • Voicing podcasts and dialogues
  • Real-time voice for assistants
Sizes
0.5B – 9B
Hardware
from: Laptop
Commercial use allowedDetails
Speech to textGGUF2025–2026

Granite Speech

IBM · USA

IBM speech models for recognizing and translating speech in English, several European languages and Japanese. Designed for enterprise use.

  • Transcribing business meetings
  • Translating speech into text in another language
  • Voice assistants
Sizes
470M – 8B
Hardware
from: Laptop
Commercial use allowedDetails
Speech to textGGUF2024–2026

Moonshine

Moonshine AI (Useful Sensors) · USA

Very small and fast speech recognition models for phones, tablets and embedded devices. Version 2 streams, producing text while the person is still speaking.

  • Voice control of devices
  • Offline recognition on a phone
  • Live subtitles
Sizes
27M – 245M
Hardware
from: Laptop
Commercial use allowedDetails
Speech to textRU2022–2026

Распознавание речи для башкирского, татарского и марийского (AigizK)

Aigiz Kunafin (AigizK), member of the SLONE community · Russia

Years of work by one enthusiast around the Bashkir Common Voice project: Whisper, wav2vec2, w2v-BERT and GigaAM fine-tuned for Bashkir, Tatar and Mari. Quality differs between models, so test them on your own recordings.

  • Transcribing calls and meetings in Bashkir and Tatar
  • Subtitles for video and radio broadcasts
  • Turning archive recordings into text
Sizes
from 220M to 1B across the different models
Hardware
from: Laptop
Commercial use with conditionsDetails
Speech to textRUGGUF2024–2026

GigaAM

Sber · Russia

Sber's models for Russian speech recognition, among the most accurate for Russian. Includes emotion recognition, v3 with punctuation, and a multilingual version (Russian, Kazakh, Kyrgyz, Uzbek).

  • Transcribing calls in Russian
  • Meeting minutes
  • Voice control of services
Sizes
220M – 600M
Hardware
from: Laptop
Commercial use allowedDetails
Speech to textGGUF2026

Cohere Transcribe

Cohere · Canada

Cohere's speech recognition model for 14 languages (Russian is not on the list), with a separate version for Arabic. Built for accurate transcription of business recordings.

  • Transcribing meetings and interviews
  • Subtitles
  • Searching an audio archive
Sizes
2B
Hardware
from: Laptop
Commercial use allowedDetails
Speech to textRUGGUF2026

Qwen3-ASR

Alibaba (Qwen) · China

Speech recognition models from the Qwen team for 50+ languages, including Russian. They handle noise, singing and accents well.

  • Transcribing calls and meetings
  • Video subtitles
  • Multilingual recognition
Sizes
0.6B – 1.7B
Hardware
from: Laptop
Commercial use allowedDetails
Speech to textRU2025

Omnilingual ASR

Meta · USA

Speech recognition for 1,600+ languages, including Russian and rare languages no system supported before. A new language can be added from a few examples.

  • Transcription in rare and local languages
  • Digitizing oral archives
  • Subtitles in many languages
Sizes
300M – 7B
Hardware
from: Laptop
Commercial use allowedDetails

Image + text 18

Image + text2026

Mage-VL-Cowork

Microsoft · USA

A compact 4B model that operates a computer by itself: it looks at a screenshot and clicks, and when that is more efficient it switches to the command line and edits files with code. Both roles are in one model, with no external orchestrator. On the OSWorld-Verified test (50 steps) it scores 66.0%, which the authors say is the best result among 4B models, and 87.5% on operating-system tasks. It was trained from the intermediate Mage-VL in five stages, with a set of almost 294 thousand synthetic tasks over 6,915 skills. The context for training and running is 32,768 tokens, and the architecture holds up to 262,144. It suits office automation and assistants on your own computer. The figures come from the authors.

  • an agent that works in applications through screenshots and the command line
  • automating office tasks on your own computer without a cloud
  • a prototype assistant that operates the desktop
Sizes
4B
Hardware
from: Laptop
Commercial use allowedDetails
Image + textOllama2024–2026

MiniCPM-V

OpenBMB (ModelBest and Tsinghua University) · China

Compact vision models that run even on a phone or laptop. Good at reading text in photos and understanding video; version 4.6 is only 1.3B. The fresh 4.7 comes in two sizes, 1B and 35B-A3B, the first MoE model in the family. According to the authors, 35B-A3B runs on an iPhone 17 Pro Max at 13+ tokens per second with 3 GB of memory thanks to their way of loading experts. The figures come from the authors themselves.

  • On-device text recognition in photos
  • Processing receipts and documents without sending them to the cloud
  • Describing photos and video
Sizes
1B – 35B-A3B
Hardware
from: Laptop
Commercial use allowedDetails
Image + textGGUF2024–2026

Ovis

Alibaba (AIDC-AI) · China

Vision models from Alibaba's international division with strong text and table reading. The line includes the Ovis2.6 MoE and separate compact OvisOCR models for documents.

  • Extracting data from invoices, contracts and delivery notes
  • Table recognition
  • Answering questions about photos and charts
Sizes
0.9B – 80B-A3B
Hardware
from: Laptop
Commercial use allowedDetails
Image + text2023–2026

InternVideo

Shanghai AI Lab (OpenGVLab) · China

A family of video models: encoders for search and classification of clips, and chat models that analyze long videos. InternVideo 3 is designed for multi-hour recordings.

  • Searching a video archive with a text query
  • Action recognition in video
  • Answering questions about a long recording
Sizes
small encoders – 9B
Hardware
from: Laptop
Commercial use allowedDetails
Image + textOllama2024–2026

Moondream

Moondream (M87 Labs) · USA

A small, fast vision model for product use cases: answering questions, finding and pointing to objects, captions. Moondream 3.1 is a 9B MoE with 2B active.

  • Finding and counting objects in photos
  • Checking photos from field reports
  • Captions and tags for a catalogue
Sizes
2B – 9B-A2B
Hardware
from: Laptop
Commercial use with conditionsDetails
Image + textGGUF2025–2026

Keye-VL

Kuaishou · China

Vision models from Kuaishou focused on short videos. Keye-VL-2.0 (30B, 3B active) understands well what happens in a clip and when.

  • Analysing and describing short videos
  • Reviewing clips and content
  • Finding the right moment in a video
Sizes
8B – 671B-A37B
Hardware
from: Laptop
Commercial use allowedDetails
Image + textOllama2023–2026

LLaVA

LLaVA / LMMs-Lab (researchers from the USA and China) · USA / China

The open project that started the trend for image-plus-text models. The OneVision line understands photos, documents and video; training data and recipes are open.

  • Answering questions about photos and screenshots
  • Describing products from a photo
  • Frame-by-frame video analysis
Sizes
0.5B – 72B
Hardware
from: Laptop
Commercial use allowedDetails
Image + textOllama2025–2026

Granite Vision

IBM · USA

Compact IBM models for business documents: tables, charts, forms, field-value pairs. The model card openly warns that it works best with English.

  • Extracting fields from forms and invoices
  • Turning charts and tables into data
  • Answering questions about documents
Sizes
2B – 4B
Hardware
from: Laptop
Commercial use allowedDetails
Image + textRUGGUF2023–2026

InternVL

Shanghai AI Laboratory (OpenGVLab) · China

A large family of Chinese vision models sized from 1B to 241B. InternVL-U (4B) combines image understanding, generation and editing.

  • Understanding documents, diagrams and charts
  • Answering questions about photos
  • Video analysis
Sizes
1B – 241B-A28B
Hardware
from: Laptop
Commercial use allowedDetails
Image + text2024–2026

Molmo

Ai2 (Allen Institute for AI) · USA

Fully open vision models from Ai2 (weights and data). They can point to a spot in an image and count objects; Molmo2 understands video, MolmoWeb controls a browser.

  • Counting products and objects in photos
  • Pointing to where an item is in an image
  • Video analysis
Sizes
1B-A7B – 72B
Hardware
from: Laptop
Commercial use allowedDetails
Image + text2023–2025

CogVLM и GLM-V

Zhipu AI (Z.ai) and Tsinghua University · China

Vision models from Zhipu: first CogVLM, then the GLM-V line. GLM-4.6V can call tools based on images and act as an agent operating an interface.

  • Answering questions about photos and documents
  • An agent that operates an interface from screenshots
  • Analysing charts and reports
Sizes
9B – 106B-A12B
Hardware
from: Laptop
Commercial use allowedDetails
Image + textRUOllama2023–2025

Qwen-VL

Alibaba (Qwen team) · China

One of the strongest open vision models: reads documents, tables, charts and video, and works with user interfaces. Since Qwen3.5, vision is built directly into the main Qwen model.

  • Extracting data from scanned invoices and delivery notes
  • Analysing photos of products and shelves
  • Analysing video and camera footage
Sizes
2B – 235B-A22B
Hardware
from: Laptop
Commercial use allowedDetails

Voice: speakers and sound 18

Voice: speakers and soundGGUF2026

Nemotron 3 Diarization

NVIDIA · USA

Marks up a recording by speaker: who spoke when, for up to eight participants, both in files and live. The model is small, ships a GGUF build and runs without a GPU. Call recordings are personal data, so retention rules are agreed with a lawyer.

  • Splitting a call or meeting recording by participant
  • Adding speaker names to a transcript
  • Live speaker labelling during a conversation
Sizes
100M
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and sound2022–2026

UVR / MDX-Net / RoFormer (разделение звука)

Community: Ultimate Vocal Remover (Anjok07), ZFTurbo, MVSep · International community

A large open collection of models for separating vocals from music and noise: MDX-Net, BS-RoFormer, Mel-RoFormer, SCNet. The quality leaders for vocals among open solutions.

  • Clean vocals from a recording with music
  • Backing tracks and stems for karaoke
  • Removing background music and noise from videos
Sizes
from tens to hundreds of millions of parameters
Hardware
from: Laptop
Commercial use with conditionsDetails
Voice: speakers and sound2022–2026

Demucs

Meta AI, then Alexandre Défossez · France

A classic model that splits a track into vocals, drums, bass and the rest. The v4 hybrid transformer version remains the benchmark; the project is now maintained by its author in his own repository.

  • Separating vocals from music in a recording
  • Backing tracks and karaoke stems
  • Cleaning speech in videos with background music
Sizes
tens of millions of parameters
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and sound2023–2026

RVC (Retrieval-based Voice Conversion)

RVC-Project community · China

The most widely used open voice conversion tool: a model for a specific voice trains on 10–30 minutes of recording and works in real time. Use only with the voice owner's consent.

  • Voicing content with one brand voice
  • Covers and vocal work
  • Real-time voice changing
Sizes
tens of millions of parameters
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and soundGGUF2022–2026

WeSpeaker

WeNet community · China

A set of ready-made voiceprint models: checks whether the same person speaks in two recordings and helps split a recording by speaker. One of the models is built into pyannote 3.x.

  • Voice verification of a customer during a call
  • Finding repeat calls from the same person
  • Splitting a recording by speaker
Sizes
from a few to tens of millions of parameters
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and soundRU2026

FireRedVAD

FireRedTeam (Xiaohongshu) · China

A speech and sound event detector: tells apart speech, singing and music. In a 102-language test (the FLEURS set, which includes Russian) it beat Silero VAD and TEN VAD. Has a streaming mode.

  • Cutting recordings before speech recognition
  • Separating speech from music and singing in broadcasts and videos
  • Speech detection in voice bots
Sizes
compact, exact size not stated
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and soundRU2025–2026

Smart Turn

Daily (Pipecat) · USA

Uses intonation to tell whether a person has finished a thought or just paused, so a voice bot does not interrupt. Version 3 is 8 MB, runs on a CPU and understands 23 languages, including Russian.

  • Voice bot does not interrupt the customer during pauses
  • Fast reply when the customer has really finished
  • An add-on to a standard speech detector in voice assistants
Sizes
8M (v3) – 580M (v1)
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and soundRU2020–2025

Silero VAD

Silero · Russia

The most popular open speech detector: tells voice apart from silence and noise. Processes an audio chunk in under a millisecond on a single CPU core; trained on recordings in more than 6,000 languages.

  • Cutting calls and recordings before speech recognition
  • Detecting when the customer is speaking in a voice bot
  • Filtering out silence and noise to save on transcription
Sizes
about 2 MB
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and sound2022–2025

NVIDIA Sortformer / TitaNet

NVIDIA · USA

NVIDIA models for "who is speaking": TitaNet recognizes a specific person's voice, Sortformer splits a recording into up to 4 speakers, including live during a call.

  • Real-time speaker tagging in conversations
  • Checking that the same person is calling (voiceprint)
  • Preparing meeting transcripts
Sizes
23M (TitaNet) – 117M (Sortformer)
Hardware
from: Laptop
Commercial use with conditionsDetails
Voice: speakers and sound2022–2025

pyannote (диаризация)

pyannoteAI (Hervé Bredin) · France

The most widely used open tool for splitting a recording by speaker: who spoke and when. Usually paired with speech recognition. Weights are issued after a short form on HF.

  • Tagging calls: which part is the agent, which is the customer
  • Meeting minutes with speaker labels
  • Preparing recordings for transcription and analysis
Sizes
a few million parameters
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and sound2024–2025

ClearerVoice (MossFormer)

Alibaba (Tongyi Lab) · China

Alibaba's set of speech cleanup models: noise suppression, separating overlapping voices, upscaling audio to 48 kHz, and isolating a voice using video of the speaker's face.

  • Noise suppression in conversation recordings
  • Separating two voices speaking at once
  • Improving old and phone recordings
Sizes
under 1B
Hardware
from: Laptop
Commercial use allowedDetails
Voice: speakers and soundRU2023–2025

Эмоции в русской речи (модели на датасете Dusha)

Community (xbgoose and others), Dusha dataset from SberDevices · Russia

Models that detect emotion from voice in Russian speech: neutral, anger, positive, sadness. Trained on the open Dusha dataset from SberDevices.

  • Finding calls with irritated customers
  • Assessing the tone of operator conversations
  • Prioritizing complaints in a call center
Sizes
21M – 316M
Hardware
from: Laptop
Commercial use with conditionsDetails

Medicine 16

Medicine2026

FleXray

MIT CSAIL, Massachusetts General Hospital and Harvard Medical School · USA

Finds 60 anatomical structures on an X-ray: ribs, vertebrae, pelvis and limb bones. They can be used to measure, for example, the curvature angle of the spine automatically. It does not look for pathologies and does not tell right from left.

  • labeling anatomy on X-ray images
  • measuring the spine curvature angle
  • matching fluoroscopy frames with CT for navigation
Sizes
a small U-Net, 256 by 256 pixel images
Hardware
from: Laptop
Non-commercial onlyDetails
MedicineGGUF2023–2026

HuatuoGPT

FreedomIntelligence (The Chinese University of Hong Kong, Shenzhen) · China

A large family of medical models: chat, an imaging version, the reasoning HuatuoGPT-o1 and the new HuatuoGPT-3 on Qwen3. Does not replace a doctor; decisions are made by a specialist.

  • Draft discharge summaries for a doctor to review
  • Searching medical literature
  • Hints for doctors when reviewing images (Vision)
Sizes
7B – 72B
Hardware
from: Laptop
Commercial use allowedDetails
MedicineOllama2023–2026

Meditron

EPFL · Switzerland

Open medical models from Swiss EPFL, fine-tuned on clinical guidelines on top of various base models. Does not replace a doctor; decisions are made by a specialist.

  • Answering staff questions based on clinical guidelines
  • Draft discharge summaries for a doctor to review
  • Searching medical literature
Sizes
2B – 70B
Hardware
from: Laptop
Commercial use with conditionsDetails
MedicineGGUF2025–2026

Hulu-Med

Zhejiang University · China

A medical model for text, images, 3D scans and video: from a light 4B to a large MoE. Does not replace a doctor; decisions are made by a specialist.

  • Hints for doctors when reviewing images and CT scans
  • Draft reports and discharge summaries
  • Searching medical literature
Sizes
4B – 235B-A22B
Hardware
from: Laptop
Commercial use allowedDetails
Medicine2023–2026

CheXagent / CheXOne

Stanford AIMI · USA

Stanford models for chest X-rays: they describe the image and prepare a draft report. Does not replace a doctor; decisions are made by a specialist.

  • A draft X-ray description for the radiologist
  • Hints for doctors when reviewing images
  • Checking reports for completeness
Sizes
3B – 8B
Hardware
from: Laptop
Commercial use with conditionsDetails
MedicineGGUF2025–2026

Lingshu

Alibaba DAMO Academy · China

Alibaba's medical model based on Qwen2.5-VL: understands many types of medical images and medical text, and can reason step by step. Does not replace a doctor; decisions are made by a specialist.

  • Hints for doctors when reviewing images
  • Draft reports and discharge summaries
  • Searching medical literature
Sizes
7B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
MedicineOllama2025–2026

MedGemma

Google · USA

Google's medical version of Gemma: reads medical texts and images (X-ray, dermatology, histology). A tool for doctors and developers; does not replace a doctor, decisions are made by a specialist.

  • Draft discharge summaries and reports for a doctor to review
  • Hints for doctors when reviewing images
  • Searching and summarising medical literature
Sizes
4B – 27B
Hardware
from: Laptop
Commercial use with conditionsDetails
MedicineGGUF2025

II-Medical

Intelligent Internet · UK

Reasoning medical models on Qwen3, designed to run on an ordinary computer. Does not replace a doctor; decisions are made by a specialist.

  • Reference answers to staff with the reasoning shown
  • Draft discharge summaries for a doctor to review
  • Searching medical literature
Sizes
7B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Medicine2025

MedSigLIP

Google · USA

Google's lightweight encoder for medical images and text, the same one inside MedGemma. Sorts images and finds similar ones. Does not replace a doctor; decisions are made by a specialist.

  • Finding similar images in a clinic's archive
  • Pre-sorting images for a doctor
  • A base for your own image classifiers
Sizes
about 0.9B
Hardware
from: Laptop
Commercial use with conditionsDetails
Medicine2024–2025

Bioptimus H-optimus

Bioptimus · France

Pathology foundation models from France's Bioptimus with 1.1B parameters, plus the compact H0-mini. The first version is open under Apache 2.0. Does not replace a doctor; decisions are made by a specialist.

  • Histology slide patch features for research models
  • A prototype for sorting slides by tissue type
  • Research projects linking morphology and molecular data
Sizes
86M (H0-mini) – 1,1B
Hardware
from: Laptop
Commercial use with conditionsDetails
Medicine2024–2025

MahmoodLab UNI / UNI 2

Mahmood Lab (Mass General Brigham, Harvard) · USA

A foundation model for histology slides: turns patches of digital slides into features for tissue classification. Does not replace a doctor; decisions are made by a specialist.

  • Research classifiers of tissue types from digital slides
  • Finding similar cases in a slide archive
  • Preparing features for research prognosis models
Sizes
about 300M (UNI) – about 680M (UNI2-h)
Hardware
from: Laptop
Non-commercial onlyDetails
Medicine2024

MahmoodLab CONCH / TITAN

Mahmood Lab (Mass General Brigham, Harvard) · USA

Image-plus-text models for pathology: search slides by an English description, classify without fine-tuning; TITAN describes a whole slide. Does not replace a doctor; decisions are made by a specialist.

  • Text-query search across a slide archive for research
  • Draft slide descriptions for research projects
  • Tissue classification without labels at the start of a study
Sizes
about 160M – 300M
Hardware
from: Laptop
Non-commercial onlyDetails

Computer vision 15

Computer vision2024–2026

MoGe

Microsoft Research · USA

Reconstructs the 3D geometry of a scene from one photo: depth in meters, a point cloud and surface normals.

  • Measuring rooms and objects from photos
  • 3D point cloud from a single shot
  • Preparing data for robots and AR
Sizes
ViT-S – ViT-G
Hardware
from: Laptop
Commercial use allowedDetails
Computer vision2025–2026

RF-DETR

Roboflow · USA

Real-time object detector, an open alternative to YOLO without AGPL. Supports segmentation (object outlines) and, since 2026, keypoints.

  • Object detection in video and photos
  • Precise outlines of parts and defects
  • Fine-tuning for your own object classes
Sizes
Nano – 2XL
Hardware
from: Laptop
Commercial use allowedDetails
Computer vision2024–2026

Sapiens

Meta · USA

Meta's models for analyzing people in photos: pose keypoints, body part segmentation, normals and depth. Sapiens2 was trained at high resolution and adds human matting.

  • Pose and body keypoint detection
  • Segmentation of body parts and clothing
  • Separating a person from the background
Sizes
0.1B – 5B
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer vision2023–2026

Segment Anything (SAM)

Meta · USA

Selects any object in photos and videos with a click or a box. The basis for background removal and object counting.

  • Background removal from product photos
  • Counting objects in photos
  • Data labeling for training
Sizes
91M – ~0,85B
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer vision2023–2026

YOLO (Ultralytics)

Ultralytics · USA

The most widely used real-time object detector: finds and marks items in video even on modest hardware. YOLOv5 came out back in 2020; the catalog starts from YOLOv8.

  • Counting people, cars and goods on video
  • Checking hard hats and workwear
  • Spotting defects on the production line
Sizes
2.4M – 68M
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer visionGGUF2024–2025

Depth Anything

ByteDance and the University of Hong Kong · China

Estimates depth, the distance to every point, from one ordinary photo or video. DA3 reconstructs scene geometry from several frames.

  • Estimating distances and volumes from a camera
  • Depth effects for photo and video
  • Navigation for robots and drones
Sizes
25M – 1.4B
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer vision2025

Perception Encoder (PE)

Meta · USA

Meta's family of encoders for images and video, and with PE-AV also for audio. PE-Core searches by text more accurately than SigLIP 2 (per Meta); small versions are available.

  • Search photos and videos by description
  • Catalog labeling and tagging
  • Search across audio and video (PE-AV)
Sizes
size not stated on the model card
Hardware
from: Laptop
Commercial use allowedDetails
Computer vision2023–2025

MetaCLIP / MetaCLIP 2

Meta · USA

Meta's open reproduction of CLIP with a transparent data collection recipe. MetaCLIP 2 is trained on multilingual data from around the world. Non-commercial license only.

  • Image search by text
  • Image classification without training
  • Search research and prototypes
Sizes
0.15B – 3.6B
Hardware
from: Laptop
Non-commercial onlyDetails
Computer vision2023–2025

Grounding DINO / Rex-Omni

IDEA Research · China

Finds any objects in an image from a text description, without training on your data: "red box", "person without a hard hat". Rex-Omni is the new VLM-based generation.

  • Finding objects by description without labeling
  • Automatic data labeling for training
  • Checking photos against requirements
Sizes
172M – 3B
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer vision2023–2025

DINOv2 / DINOv3

Meta · USA

Foundation models that turn an image into a numeric "fingerprint". They are used to build similar-image search, classification and segmentation without large labeled datasets.

  • Finding similar products and photos
  • Image classification on small datasets
  • Base for your own quality-control models
Sizes
21M – 7B
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer vision2023–2025

SigLIP (наследник CLIP)

Google · USA

Models that map images and text into a shared space: you can search photos by words and classify images without training. OpenAI's CLIP (2021) is the predecessor.

  • Image search by text query
  • Automatic catalog labeling and tagging
  • Filtering prohibited content
Sizes
about 0.2B to 2B
Hardware
from: Laptop
Commercial use allowedDetails
Computer vision2022–2025

ViTPose

University of Sydney and JD Explore Academy · Australia / China

A simple, accurate model for human pose estimation via keypoints. ViTPose++ handles human, animal and whole-body poses; built into the Transformers library.

  • Body keypoints in photos and video
  • Motion analysis in sports and rehabilitation
  • Monitoring work postures and safety practices
Sizes
33M – about 1B
Hardware
from: Laptop
Commercial use allowedDetails

Code 15

Code2026

FrogNano

Microsoft · USA

A compact agent for working with repository code, based on Qwen3.5-4B: searches the project, edits files, runs tests and prepares a patch. Trained only with reinforcement learning on about 1500 synthetic tasks. Works through the light Leaf harness with five tools.

  • fixing bugs in a repository from a task description
  • navigating an unfamiliar codebase
  • preparing patches with test runs
Sizes
4B
Hardware
from: Laptop
Commercial use allowedDetails
CodeGGUF2025–2026

Mellum

JetBrains · Czech Republic

JetBrains models for fast code autocompletion. Mellum2 (12B, 2.5B active) is already a full assistant: it writes and edits code, calls tools and reasons. Version 2.1 spent most of its training on reinforcement learning in real repositories with a shell and tests: 47 on SWE-bench Verified, self-reported by JetBrains.

  • Fast code autocompletion on your own server
  • A developer assistant that does not send code to the cloud
  • Fine-tuning on the company's code
Sizes
4B – 12B-A2.5B
Hardware
from: Laptop
Commercial use allowedDetails
CodeOllama2025–2026

Rnj-1

Essential AI · USA

An 8B model trained from scratch by the company of one of the authors of the transformer architecture. Strong at code and technical tasks; version 1.5 handles context up to 160K tokens.

  • Writing and fixing code
  • A developer agent on a single GPU
  • Solving technical and scientific problems
Sizes
8B
Hardware
from: Laptop
Commercial use allowedDetails
CodeGGUF2026

IQuest-Coder

IQuest Research · China

A family of coding models with standard and reasoning versions, including a Loop variant that runs through its layers a second time. Sizes from 7B to 40B.

  • Writing and refining code
  • Solving tasks with step-by-step reasoning
  • Agentic work with a repository
Sizes
7B – 40B
Hardware
from: Laptop
Commercial use with conditionsDetails
CodeOllama2024–2026

Qwen Coder

Alibaba (Qwen team) · China

The broadest open coding family: from 0.5B for autocompletion to 480B for agents. Qwen3-Coder-Next (80B, 3B active) works as a developer agent on a single GPU.

  • Code autocompletion in the editor
  • An agent that edits code in the repository on its own
  • Writing and refining scripts, SQL and integrations
Sizes
0.5B – 480B-A35B
Hardware
from: Laptop
Commercial use allowedDetails
Code2026

SERA (Ai2 Open Coding Agents)

Ai2 (Allen Institute for AI) · USA

Fully open developer agents from Ai2: weights, data and training recipe are all public. Designed so a company can cheaply fine-tune the agent on its own repository.

  • An agent for fixing issues in code
  • Fine-tuning the agent on an internal repository
  • Automating small edits and tests
Sizes
8B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
CodeRU2025

Kodify-Nano (МТС AI)

MTS AI (MWS AI) · Russia

A small coding assistant from MTS AI that understands requests in Russian. Runs locally, with plugins for VS Code and JetBrains.

  • Code suggestions and completion in the editor
  • Code explanations in Russian
  • Drafts of tests and documentation
Sizes
1.5B
Hardware
from: Laptop
Commercial use allowedDetails
CodeGGUF2025

Seed-Coder

ByteDance Seed · China

A compact 8B coding model from ByteDance in base, instruct and reasoning versions. Its training data was selected by the model itself, with almost no hand-written rules.

  • Code autocompletion and generation
  • Solving algorithmic problems
  • A base for fine-tuning on your own stack
Sizes
8B
Hardware
from: Laptop
Commercial use allowedDetails
CodeOllama2024

OpenCoder

INF Technology · China

Fully reproducible coding models: along with the weights, the data, its cleaning pipeline and the training recipe are open. Understand English and Chinese.

  • Code generation and completion
  • Training your own coding model from an open recipe
  • A programming assistant on low-end hardware
Sizes
1.5B – 8B
Hardware
from: Laptop
Commercial use with conditionsDetails
CodeOllama2023–2024

DeepSeek-Coder

DeepSeek · China

DeepSeek's coding model family: from small autocompletion models to the large MoE V2, which matched closed models in 2024. Later, coding moved into DeepSeek's general models.

  • Code autocompletion and generation
  • Translating code between programming languages
  • Finding bugs and explaining other people's code
Sizes
1.3B – 236B-A21B
Hardware
from: Laptop
Commercial use with conditionsDetails
CodeOllama2024

Yi-Coder

01.AI · China

Coding models from 01.AI at 1.5B and 9B with a 128K-token context and support for 52 programming languages. A separate line next to the text Yi models.

  • Code autocompletion and generation
  • Explaining and refactoring code
  • A programming assistant without the cloud
Sizes
1.5B – 9B
Hardware
from: Laptop
Commercial use allowedDetails
CodeOllamaNot maintained2023–2024

CodeGeeX

Zhipu AI (Z.ai) and Tsinghua University · China

Coding models from the creators of GLM. CodeGeeX4-ALL-9B, based on GLM-4-9B, combines autocompletion, code chat, function calling and repository search in one model.

  • Code autocompletion in the IDE
  • A code chat assistant
  • Answering questions about a repository
Sizes
6B – 9B
Hardware
from: Laptop
Commercial use with conditionsDetails

Deepfake detection 15

Deepfake detection2023–2026

TrustMark

Adobe Research and University of Surrey · USA

An image watermark for arbitrary resolutions built for the Content Authenticity Initiative: it can both apply a mark and remove one. The detector errs in both directions - a human reviews the output.

  • Marking images on the way out of your own pipeline
  • Checking the provenance of a submitted image
  • Linking with content provenance metadata
Sizes
model types Q and P with different mark capacity
Hardware
from: Laptop
Commercial use allowedDetails
Deepfake detection2025–2026

Community Forensics

University of Michigan · USA

A lightweight detector of generated images, trained on 2.7M samples from nearly 5000 different generators. It errs in both directions: the result is a reason for a human to check, not proof.

  • Checking submitted photos and illustrations
  • Filtering AI images in a content flow
  • Flagging suspicious images for manual review
Sizes
22M
Hardware
from: Laptop
Commercial use allowedDetails
Deepfake detection2023–2026

UniversalFakeDetect

University of Wisconsin-Madison · USA

An early and still used approach: a simple classifier trained on top of a frozen CLIP that transfers to unseen generators. It errs in both directions - the output needs a human check.

  • Checking images from new, unfamiliar generators
  • A baseline when comparing detectors
  • Fast rollout of a check without training a large model
Sizes
a linear classifier on top of CLIP ViT-L/14
Hardware
from: Laptop
Commercial use allowedDetails
Deepfake detection2024–2025

VideoSeal

Meta · USA

A watermark for video and images that survives re-encoding and cropping. The detector errs in both directions: a missing mark does not prove a forgery, and finding one is a reason for a human to check.

  • Marking video created or processed by AI
  • Finding your own mark in re-uploaded clips
  • Protecting ad materials from being reused as someone else's
Sizes
a mark of 96 to 1024 bits
Hardware
from: Laptop
Commercial use allowedDetails
Deepfake detection2025

AntiDeepfake (NII)

National Institute of Informatics, Yamagishi Lab · Japan

Seven speech encoders (wav2vec 2.0, XLS-R, MMS, HuBERT) post-trained to tell live speech from synthetic. The authors note themselves that quality depends heavily on the dataset; a human reviews the output.

  • Checking audio recordings for synthesis
  • Fine-tuning for your own language and recording channel
  • Comparing several encoders on your own data
Sizes
0,3B – 2B
Hardware
from: Laptop
Non-commercial onlyDetails
Deepfake detection2023–2025

RADAR

IBM Research and The Chinese University of Hong Kong · USA

An AI-text detector trained together with a paraphraser: it was deliberately taught not to give up when the text has been rewritten. It errs in both directions; a human reviews the output.

  • Checking texts that may have been rewritten after generation
  • First-pass filtering in a newsroom or admissions office
  • Comparison against simpler detectors
Sizes
about 355M (RoBERTa-large)
Hardware
from: Laptop
Commercial use with conditionsDetails
Deepfake detection2023–2025

DeepfakeBench

The Chinese University of Hong Kong, Shenzhen (SCLBD) · China

Dozens of open face-swap detectors for video and photo under one codebase with ready weights. A detector errs in both directions: its output is a reason for a human to check, not proof of a forgery.

  • First-pass check of a submitted video or selfie
  • Comparing several detectors on your own data
  • Fine-tuning a detector for your own flow of applications
Sizes
Xception- and EfficientNet-class detectors, tens of millions of parameters
Hardware
from: Laptop
Non-commercial onlyDetails
Deepfake detection2024–2025

Watermark Anything (WAM)

Meta · USA

An image watermark that can be applied to individual regions: the model shows which part of the image is marked. It errs in both directions - a human reviews the result.

  • Marking generated and edited images
  • Finding a marked fragment inside a collage
  • Tracking which parts of a picture were made by AI
Sizes
a mark encoder and decoder for images
Hardware
from: Laptop
Commercial use allowedDetails
Deepfake detection2025

Desklib AI Text Detector

Desklib · India

A recent open AI-text detector on DeBERTa-v3-large, trained on the RAID dataset, with a separate version for academic work. It errs in both directions - a human always reviews the result.

  • Checking submitted articles and reports
  • Filtering templated reviews and applications
  • First-pass check of student work
Sizes
0.4B (DeBERTa-v3-large)
Hardware
from: Laptop
Commercial use allowedDetails
Deepfake detection2024

AudioSeal

Meta · USA

An imperceptible mark in synthetic speech plus a fast detector that finds it even inside a fragment of a long recording. The detector errs in both directions: a hit is a reason for a human to check, not proof.

  • Marking speech synthesized by your service
  • Finding your own mark in third-party publications
  • Checking whether synthesis was mixed into a call recording
Sizes
a watermark generator and detector, 16-bit message
Hardware
from: Laptop
Commercial use allowedDetails
Deepfake detectionNot maintained2024

MAGE

UC Santa Barbara and co-authors · USA

A Longformer-based AI-text detector: it holds a long document whole and was trained on texts from many different language models. It errs in both directions; its output is a reason for a human to check.

  • Checking long articles and reports as a whole
  • Filtering machine text in a publication flow
  • Comparing detectors on your own data
Sizes
about 150M (Longformer-base)
Hardware
from: Laptop
Commercial use allowedDetails
Deepfake detectionNot maintained2023

HC3 ChatGPT Detector

Hello-SimpleAI · China

One of the first open AI-text classifiers, trained on the HC3 corpus of paired human and ChatGPT answers. It errs in both directions: its output is a reason to talk to the author, not proof.

  • First-pass check of student work
  • Filtering templated applications and reviews
  • Flagging suspicious texts for manual review
Sizes
about 125M (RoBERTa-base)
Hardware
from: Laptop
Commercial use with conditionsDetails

Photo editing 14

Photo editing2023–2026

BRIA RMBG

BRIA AI · Israel

BRIA's background removal, trained on licensed photos. Soft edges, hair, transparency. Video versions available. Business use requires a paid agreement.

  • Cutting products out onto a white background
  • Staff and expert photos without background
  • Background removal in video
Sizes
44M – 220M
Hardware
from: Laptop
Non-commercial onlyDetails
Photo editing2025–2026

MatAnyone

S-Lab, Nanyang Technological University · Singapore

Cuts a person out of video with a precise alpha mask, including hair and edges, without a green screen. Needs a first-frame mask, for example from SAM.

  • Background replacement in video without chroma key
  • Cutting out a person for editing and effects
  • Preparing videos for advertising and social media
Sizes
about 35M
Hardware
from: Laptop
Non-commercial onlyDetails
Photo editingGGUF2024–2025

BiRefNet

Nankai University · China

An open MIT-licensed model for precise object segmentation and background removal. RMBG-2.0 is built on it. Versions for 2K and for hair and semi-transparent edges.

  • Bulk background removal from product photos
  • Precise masks for design and print
  • Cutting out people with hair for advertising
Sizes
about 220M (lightweight lite versions available)
Hardware
from: Laptop
Commercial use allowedDetails
Photo editing2024–2025

BEN2

Prama LLC · USA

A background removal model focused on difficult edges: hair, fur, fine details. The open version is MIT-licensed and can process video.

  • Cutting out products and people from photos
  • Background removal in video
  • Preparing photos for a catalog
Sizes
about 95M
Hardware
from: Laptop
Commercial use allowedDetails
Photo editingNot maintained2024

IC-Light

Lvmin Zhang (author of ControlNet) · USA

Changes lighting in a photo: relights an object or person from a description or to match a given background, so a cut-out looks natural.

  • Matching product lighting to a new background
  • Studio lighting for portraits without a reshoot
  • Consistent lighting style across a catalog
Sizes
based on SD 1.5
Hardware
from: Laptop
Commercial use allowedDetails
Photo editingNot maintained2023–2024

PowerPaint

Shanghai AI Laboratory (OpenMMLab) and Tsinghua University · China

All-round photo inpainting: remove an object, insert a new one from a description, change a shape or extend the frame beyond its edges.

  • Removing and replacing objects in photos
  • Extending the frame to a required format
  • Inserting a product or detail from a text description
Sizes
based on SD 1.5
Hardware
from: Laptop
Commercial use allowedDetails
Photo editingNot maintained2023

DDColor

Alibaba DAMO Academy · China

Colorizes black-and-white photos in natural colors. A lightweight model with a commercial-friendly license; a compact tiny version is available.

  • Colorizing archival photos
  • Color versions of historical photos for publications
  • Family photo restoration service
Sizes
DDColor-T (tiny) and DDColor-L
Hardware
from: Laptop
Commercial use allowedDetails
Photo editingNot maintained2022–2023

InSPyReNet / transparent-background

Taehoon Kim (POSTECH) · South Korea

A salient object detection model and the ready-made transparent-background tool built on it: removes backgrounds from photos, video and webcam with one command.

  • Batch background removal from photos
  • Replacing the background with a color or blur
  • Background removal in video
Sizes
small (based on Swin-B)
Hardware
from: Laptop
Commercial use allowedDetails
Photo editingNot maintained2022–2023

HAT

XPixel Group (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, and others) · China

Transformer-based photo upscaling, more accurate than SwinIR on fine details. Versions for real noisy photos and a lightweight HAT-S.

  • Upscaling product and interior photos
  • Preparing images for print
  • Sharpening archival photos
Sizes
9M – 40M
Hardware
from: Laptop
Commercial use allowedDetails
Photo editingNot maintained2022–2023

CodeFormer

S-Lab, Nanyang Technological University · Singapore

Popular face restoration for old and blurry photos; also works on video. Has face inpainting and colorization modes. Non-commercial license.

  • Restoring faces in old photos
  • Enhancing faces in low-quality video
  • Family archive restoration pilots
Sizes
small, up to 0.1B
Hardware
from: Laptop
Non-commercial onlyDetails
Photo editingGGUFNot maintained2021–2022

Real-ESRGAN

Tencent ARC Lab · China

The classic for upscaling photos 2–4x while cleaning noise and compression artifacts. Lightweight, runs even on a CPU. Versions for drawings and anime.

  • Upscaling old and small product photos
  • Cleaning images of compression artifacts
  • Preparing images for print
Sizes
about 17M
Hardware
from: Laptop
Commercial use allowedDetails
Photo editingNot maintained2021–2022

GFPGAN

Tencent ARC Lab · China

Proven face restoration for old and compressed photos, with a commercial-friendly license. Often paired with Real-ESRGAN; the most used versions are 1.3 and 1.4.

  • Restoring faces in old photos
  • Enhancing avatars and profile photos
  • Restoration in a photo shop or online service
Sizes
small, up to 0.1B
Hardware
from: Laptop
Commercial use allowedDetails

Math and reasoning 12

Math and reasoningOllama2025–2026

OpenThinker

Open Thoughts (Stanford, Berkeley and other universities) · USA

Fully open reasoning models: both weights and training data are published. Newer OpenThinkerAgent versions can carry out multi-step tasks.

  • Calculations and formula checks
  • Complex analytics with step-by-step breakdowns
  • Checking the logic of internal policies
Sizes
1.5B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Math and reasoningGGUF2025–2026

Goedel-Prover

Princeton University · USA

Open models for formal proofs in Lean 4 from Princeton. The new Goedel-Code-Prover proves program correctness.

  • Formal verification of mathematical workings
  • Verifying code correctness
  • Training
Sizes
7B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Math and reasoningGGUF2026

QED-Nano

LM Provers (CMU, Hugging Face, ETH Zurich, Project Numina) · USA, Switzerland, France

A small 4B model on Qwen3 that writes mathematical proofs in plain language almost at the level of large models. Runs on a laptop.

  • Checking the logic of reasoning and workings
  • Step-by-step explanations of solutions
  • Training and olympiad preparation
Sizes
4B
Hardware
from: Laptop
Commercial use allowedDetails
Math and reasoningGGUF2024–2025

DeepSeek-Math

DeepSeek · China

DeepSeek's maths models. The first 7B version introduced the GRPO training method; the 685B V2 writes and checks its own olympiad-level proofs.

  • Calculations and formula checks
  • Checking mathematical workings in reports
  • Working through problems step by step
Sizes
7B – 685B
Hardware
from: Laptop
Commercial use with conditionsDetails
Math and reasoningGGUF2025

Kimina-Prover

Moonshot AI and Project Numina · China, France

Models for formal proofs in Lean 4 from Moonshot AI (Kimi) and Numina. Small versions from 0.6B run on a laptop.

  • Formal verification of mathematical workings
  • Translating a problem from plain language into Lean
  • Training and olympiad preparation
Sizes
0.6B – 72B
Hardware
from: Laptop
Commercial use allowedDetails
Math and reasoningOllama2025

DeepScaleR, DeepCoder, DeepSWE

Agentica (Berkeley, Sky Computing Lab) and Together AI · USA

Small models fine-tuned with reinforcement learning: DeepScaleR (1.5B) solves olympiad maths, DeepCoder writes code, DeepSWE works as a developer agent. Recipes and data are open.

  • Solving maths problems with step-by-step working
  • Generating and checking code
  • An agent for fixing bugs in a repository
Sizes
1.5B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Math and reasoning2025

AceMath / AceReason

NVIDIA · USA

NVIDIA models for maths and reasoning based on Qwen. AceReason was fine-tuned with reinforcement learning first on maths, then on code.

  • Calculations and formula checks
  • Complex analytics with step-by-step breakdowns
  • Working through programming problems
Sizes
1.5B – 72B
Hardware
from: Laptop
Commercial use with conditionsDetails
Math and reasoningGGUF2024–2025

DeepSeek-Prover

DeepSeek · China

DeepSeek models for formal proofs in Lean 4: the proof is checked by a program, not a person. A narrow tool for mathematicians and engineers.

  • Formal verification of mathematical workings
  • Verifying algorithm correctness
  • Training and olympiad preparation
Sizes
7B – 671B
Hardware
from: Laptop
Commercial use with conditionsDetails
Math and reasoningGGUF2025

Light-R1

Qihoo 360 · China

Reasoning models from Qihoo 360: a standard Qwen2.5 was fine-tuned for long reasoning using an open recipe; data and code are published.

  • Calculations and formula checks
  • Working through problems step by step
  • A base for your own reasoning fine-tuning
Sizes
7B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Math and reasoningGGUF2025

s1

Stanford University · USA

A reasoning model trained on just a thousand problems. It can be told to think longer to answer a hard question more accurately.

  • Calculations and formula checks
  • Working through complex problems step by step
  • Training
Sizes
1.5B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Math and reasoningGGUF2025

Sky-T1

NovaSky (Sky Computing Lab, Berkeley) · USA

A Berkeley reasoning model trained for under 450 dollars. It showed that o1-preview-level reasoning can be reproduced with modest resources.

  • Calculations and formula checks
  • Working through problems step by step
  • A base for your own reasoning fine-tuning
Sizes
7B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Math and reasoningOllama2024–2025

Qwen2.5-Math

Qwen (Alibaba) · China

Maths versions of Qwen: they solve problems step by step and can calculate via code. Includes reward models that check each step of a solution.

  • Calculations and formula checks
  • Checking calculations in estimates and reports
  • Working through problems step by step
Sizes
1.5B – 72B
Hardware
from: Laptop
Commercial use with conditionsDetails

Forecasting 12

Forecasting2026

Timestral

Mistral AI · France

A family of models for forecasting time series with no task-specific training: from the history of sales, demand or load it gives a forecast with intervals (21 quantiles). It takes up to 32 thousand points of history and forecasts up to 1024 steps at a time. By Mistral's measurements, the best zero-shot model on GIFT-Eval, TIME and fev-bench. Runs on a CPU.

  • forecasting sales and demand from history
  • forecasting load on a warehouse, a till, a call center
  • assessing the risk of stock shortage or surplus from forecast intervals
Sizes
5M, 28M, 123M and 209M
Hardware
from: Laptop
Commercial use with conditionsDetails
Forecasting2024–2026

TimesFM

Google · USA

A ready-made Google forecasting model: forecasts any time series without training on your data.

  • Sales and demand forecasting
  • Purchase and inventory planning
  • Load and traffic forecasting
Sizes
200M – 500M
Hardware
from: Laptop
Commercial use with conditionsDetails
Forecasting2025–2026

TiRex

NXAI · Austria

A compact forecasting model on the xLSTM architecture, a leader in open benchmarks despite its small size. Runs fast on a regular CPU.

  • Demand and sales forecasting
  • Energy consumption forecasting
  • Forecasts on modest hardware and on site
Sizes
about 35M to 82M
Hardware
from: Laptop
Commercial use allowedDetails
Forecasting2024–2026

Timer (THUML)

THUML, Tsinghua University · China

A compact forecasting foundation model from the Tsinghua lab: trained on a large set of diverse series and fine-tunable on your own data.

  • Forecasting demand and load
  • Forecasting sensor readings on the shop floor
  • Fine-tuning forecasts on your own history
Sizes
84M (timer-base)
Hardware
from: Laptop
Commercial use allowedDetails
ForecastingGGUF2025–2026

Toto

Datadog · USA

A Datadog forecasting model trained on server and application metrics. Especially strong for IT monitoring: load, latency, errors.

  • Server load forecasting
  • Anomaly detection in metrics
  • Capacity planning
Sizes
4M – 2.5B
Hardware
from: Laptop
Commercial use allowedDetails
ForecastingGGUF2024–2025

Chronos

Amazon · USA

Amazon forecasting models, among the most downloaded. Chronos-2 takes external factors into account: prices, promotions, weather.

  • Demand forecasting with promotions and prices
  • Inventory planning
  • Forecasting revenue and customer flow
Sizes
8M – 710M
Hardware
from: Laptop
Commercial use allowedDetails
ForecastingGGUF2024–2025

Moirai

Salesforce · USA

Salesforce's universal forecasting model for series with different frequencies and many variables. Weights are open for research only.

  • Research forecasting pilots
  • Comparison with other forecasting models
  • Forecasts across many related series
Sizes
11M – 311M
Hardware
from: Laptop
Non-commercial onlyDetails
Forecasting2025

Sundial

THUML, Tsinghua University · China

A forecasting model that returns a set of possible scenarios rather than a single line — useful when you need a range for demand or load, not one number.

  • Forecasting demand with a range of values
  • Planning stock while accounting for spread
  • Forecasting load on services and staff
Sizes
128M (sundial-base)
Hardware
from: Laptop
Commercial use allowedDetails
Forecasting2024

Granite Time Series (TinyTimeMixers, PatchTST)

IBM Research · USA

Tiny forecasting models from IBM: they run on an ordinary CPU and sit next to the business system without a separate GPU server.

  • Forecasting sales and warehouse stock
  • Forecasting energy use and equipment load
  • Fast forecasts right on the company server
Sizes
very small: TinyTimeMixers have about 1M parameters
Hardware
from: Laptop
Commercial use allowedDetails
Forecasting2024

MOMENT

Auton Lab, Carnegie Mellon University · USA

A foundation model for numeric series: one engine is used for forecasting, anomaly detection, filling gaps and classification.

  • Forecasting demand and load
  • Detecting anomalies in sensor readings and metrics
  • Filling gaps in historical data
Sizes
about 40M – 385M
Hardware
from: Laptop
Commercial use allowedDetails
Forecasting2024

Time-MoE

The Time-MoE team · not disclosed

A forecasting model with a sparse architecture: only part of the network runs at each step, so it stays fast at a small size.

  • Forecasting sales and stock levels
  • Forecasting load on services and staff
  • Planning purchases from history
Sizes
50M and 200M
Hardware
from: Laptop
Commercial use allowedDetails
ForecastingNot maintained2024

Lag-Llama

ServiceNow, Mila and partners · Canada

One of the first open out-of-the-box forecasting models. Tiny, gives a probabilistic forecast, now behind Chronos and TimesFM.

  • Probabilistic sales forecast
  • Quick forecasting pilots
  • Baseline model for comparison
Sizes
2.4M
Hardware
from: Laptop
Commercial use allowedDetails

Acoustic monitoring 12

Acoustic monitoring2019–2026

YAMNet

Google · USA

A tiny model covering 521 sound events, including alarms, breaking glass and screams: it fits on a microcontroller and runs without a GPU. A microphone can capture people voices, which is personal data, so check the procedure with a lawyer.

  • Fast sound labelling right on the device
  • Detecting alarm sounds on a site
  • Picking interesting fragments out of a continuous recording
Sizes
3.7M
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2021–2026

MLPerf Tiny Anomaly Detection

MLCommons · USA

A reference autoencoder for finding anomalies in machine sound: it learns only from recordings of a healthy unit and flags deviations. The system does not make a diagnosis, it gives you a reason to check the unit before it fails.

  • Listening to a machine tool, pump or conveyor
  • Flagging deviations from a unit usual noise
  • Training on your own recordings of healthy equipment
Sizes
a tiny fully connected autoencoder sized for microcontrollers
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2023–2026

BirdNET

Cornell Lab of Ornithology and Chemnitz University of Technology · USA and Germany

Recognition of more than 6000 bird species by voice, the baseline tool for acoustic monitoring of an area. A microphone on site also records people voices, which is personal data, so check the procedure with a lawyer.

  • Long-term monitoring of the sound background of a site
  • Assessing biodiversity for an environmental review
  • Selecting events from round-the-clock recordings
Sizes
a compact EfficientNet-B0 based model
Hardware
from: Laptop
Non-commercial onlyDetails
Acoustic monitoring2019–2026

PANNs

University of Surrey · UK

The classic set of convolutional networks that label sound across the 527 AudioSet categories, from machinery noise and alarms to breaking glass and screams. The system does not make a diagnosis, it gives you a reason to check the unit before it fails. A microphone can also capture people voices, which is personal data, so check the procedure with a lawyer.

  • Labelling what is happening in a microphone recording
  • Detecting alarms, breaking glass, screams
  • A base for your own model tuned to one shop floor
Sizes
from 5.9M (CNN6) to 81.9M (CNN14)
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2023–2026

CED

Xiaomi · China

Compact sound-labelling models from 5.5M to 86M, with ONNX and INT8 builds that run on an ordinary CPU and on a board next to the equipment. The system does not make a diagnosis, it gives you a reason to check the unit before it fails.

  • Labelling sound on modest hardware and on site
  • Detecting alarm sounds and abnormal noise
  • Fine-tuning for your own set of equipment sounds
Sizes
5.5M - 86M
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2024–2026

Dasheng

Xiaomi · China

A general-purpose audio encoder trained on 272 thousand hours of speech, music and noise: it gives features on top of which you train your own classifier of abnormal sounds. The system does not make a diagnosis, it gives you a reason to check the unit before it fails.

  • Audio features for your own anomaly model
  • Finding similar fragments in a recording archive
  • Fine-tuning for the sounds of one production line
Sizes
86M - 1.2B
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2025

OpenBEATs

Carnegie Mellon University and co-authors (the ESPnet project) · USA

A fully open reproduction of BEATs: code, training recipes and weights are all published. Teams pick it when they need a transparent base for their own sound model. The system does not make a diagnosis, it gives you a reason to check the unit before it fails.

  • A transparent base for your own audio model
  • Fine-tuning on your own equipment recordings
  • Labelling sound events
Sizes
base and large versions, exact sizes are listed on the model cards
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2025

Perch

Google DeepMind · USA

A model for acoustic monitoring of nature: it recognises about 15,000 species and provides features for your own tasks, and the license allows commercial use. A microphone on site also records people voices, which is personal data, so check the procedure with a lawyer.

  • Monitoring the sound background of a site or water area
  • Environmental audio features for your own model
  • Selecting events from round-the-clock recordings
Sizes
size not stated on the model card
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2024–2025

EAT

Shanghai Jiao Tong University and Peng Cheng Laboratory · China

A self-supervised sound understanding model that is markedly cheaper to train on your own data than its predecessors. The system does not make a diagnosis, it gives you a reason to check the unit before it fails.

  • Training your own audio model on modest hardware
  • Audio features for spotting abnormal operating modes
  • Labelling sound events on a site
Sizes
90M (base) and 309M (large)
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2024–2025

DASS

MIT CSAIL and MIT-IBM Watson AI Lab · USA

A sound understanding model built on state spaces instead of a transformer: it handles recordings hours long, which suits continuous listening to a production line. The system does not make a diagnosis, it gives you a reason to check the unit before it fails.

  • Processing long continuous recordings
  • Finding a rare event across a multi-hour shift
  • Labelling sound events on a site
Sizes
30M (small) and 49M (medium)
Hardware
from: Laptop
Commercial use allowedDetails
Acoustic monitoring2024–2025

HeAR

Google · USA

A model for health sounds: coughing, breathing, throat clearing. It provides features on top of which a researcher builds their own model and draws no conclusions about illness itself. It does not replace a doctor; decisions are made by a specialist. Voice recordings are personal data, so check the procedure with a lawyer.

  • Features of cough and breathing sounds for your own model
  • Research projects in health acoustics
  • Selecting cough and breathing fragments from a recording
Sizes
a ViT-Large class model, trained on more than 300 million two-second clips
Hardware
from: Laptop
Commercial use with conditionsDetails
Acoustic monitoringNot maintained2022

BEATs

Microsoft · USA

A self-supervised sound understanding model that became the base for many current systems: teams take it as a starting point and fine-tune it on their own equipment sounds. The system does not make a diagnosis, it gives you a reason to check the unit before it fails.

  • A base for your own sound understanding model
  • Audio features for spotting abnormal operating modes
  • Labelling sound events on a site
Sizes
a ViT-base class model, around 90M parameters
Hardware
from: Laptop
Commercial use allowedDetails

Plates and vehicles 12

Plates and vehicles2020–2026

BoxMOT

Mikel Brostrom (open project) · not disclosed

A set of trackers: it links the same vehicle across frames and cameras, which is how flow and time on site are counted. Filming vehicles and people and storing the records is personal data processing: agree the procedure with a lawyer.

  • Counting vehicles crossing a line
  • Following a vehicle across frames
  • Finding the same vehicle on another camera
Sizes
trackers plus re-identification models from compact to large
Hardware
from: Laptop
Commercial use with conditionsDetails
Plates and vehicles2026

Intel Vehicle Models (OpenVINO)

Intel · USA

Intel models for vehicle video analytics: car, truck, bus, motorcycle, plus entry and exit logging and vehicles stopped too long. Filming vehicles and people is personal data processing: agree the procedure with a lawyer.

  • Counting vehicles and splitting them by type
  • Logging vehicle entry and exit
  • Monitoring parking space occupancy
Sizes
YOLO26 variants from nano to x
Hardware
from: Laptop
Commercial use allowedDetails
Plates and vehicles2024–2026

Open Image Models

ankandrew (open project) · not disclosed

Ready-made ONNX detectors that find the license plate in a frame, plus general object detectors. Filming vehicles and people and storing the records is personal data processing: agree the procedure with a lawyer.

  • Finding the plate in a camera frame
  • Preparing a crop for character reading
  • Picking frames with vehicles out of a video stream
Sizes
YOLO-v9 detectors at input sizes from 256 to 640 pixels
Hardware
from: Laptop
Commercial use allowedDetails
Plates and vehicles2025–2026

YOLOv11 License Plate Detection

MorseTechLab · not disclosed

The most downloaded plate detector on Hugging Face: it finds the plate in a frame, and a separate model then reads it. Filming vehicles and people and storing the records is personal data processing: agree the procedure with a lawyer.

  • Finding the plate in camera video
  • Preparing crops for plate reading
  • Testing a plate recognition idea without training
Sizes
a fine-tuned YOLO11, with an ONNX build available
Hardware
from: Laptop
Commercial use with conditionsDetails
Plates and vehicles2024–2026

FastALPR

ankandrew (open project) · not disclosed

A ready pipeline for plate recognition: plate detector and character reading in one package, with your own models pluggable. Filming vehicles and people and storing the records is personal data processing: agree the procedure with a lawyer.

  • Getting plate recognition running quickly on site
  • Automatic vehicle access by list
  • Logging vehicle entry and exit
Sizes
a lightweight plate detector paired with XS and S reading models
Hardware
from: Laptop
Commercial use allowedDetails
Plates and vehicles2024–2026

FastPlateOCR

ankandrew (open project) · not disclosed

Lightweight models that read characters from an already cropped license plate and run on an ordinary CPU. Filming vehicles and people and storing such records is personal data processing: agree the procedure and retention with a lawyer.

  • Reading a plate at entry and exit
  • Matching a plate against a whitelist of passes
  • Going through archived camera footage
Sizes
XS and S models on the CCT architecture
Hardware
from: Laptop
Commercial use allowedDetails
Plates and vehicles2022–2026

PP-Vehicle (PaddleDetection)

PaddlePaddle (Baidu) · China

A ready set of models for vehicle tasks: plate recognition, car color and type, counting entries and exits, spotting parking where it is not allowed. Filming vehicles and people is personal data processing: agree the procedure with a lawyer.

  • Counting traffic flow
  • Identifying car type and color
  • Reading the plate at the entrance
Sizes
a set of models: vehicle detector, attributes, plate reading on PP-OCR
Hardware
from: Laptop
Commercial use allowedDetails
Plates and vehicles2026

Autolane ALPR

Autolane · not disclosed

A recent set of plate recognition weights: two plate detectors and a reading model, trained on US plates, so the Russian format needs fine-tuning. Filming vehicles and people is personal data processing: agree the procedure with a lawyer.

  • Finding the plate in a frame
  • Reading the plate characters
  • Running on cameras and single-board computers
Sizes
RF-DETR and YOLOX-S detectors, CCT-S for character reading
Hardware
from: Laptop
Commercial use allowedDetails
Plates and vehicles2023–2025

Hezar License Plate Recognition

Hezar AI · Iran

Character reading for Iranian license plates; the model is small and runs without a GPU. It was trained on another country plates, so the Russian format needs fine-tuning. Filming vehicles and people is personal data processing: agree the procedure with a lawyer.

  • Reading a plate from a cropped image
  • Using the architecture as an example for your own plate format
  • Fine-tuning reading on your own footage
Sizes
a compact CRNN model
Hardware
from: Laptop
Commercial use with conditionsDetails
Plates and vehicles2023–2025

HyperLPR

szad670401 (open project) · China

A well-known open system for Chinese plates: single-row blue and yellow plates, electric vehicles, driving-school cars. It has to be fine-tuned for the Russian plate format. Filming vehicles and people is personal data processing: agree the procedure with a lawyer.

  • Plate recognition at entry and exit
  • Running on modest hardware without a GPU
  • Embedding recognition into your own application
Sizes
compact detection and reading models designed to run without a GPU
Hardware
from: Laptop
Commercial use allowedDetails
Plates and vehiclesRU2022–2024

Nomeroff Net

RIA.com · Ukraine

A plate recognition system with separate reading modules for Russia, Kazakhstan, Belarus, Ukraine, Georgia, Armenia and Kyrgyzstan. Filming vehicles and people and storing the records is personal data processing: agree the procedure with a lawyer.

  • Reading a Russian plate from a camera
  • Identifying the country and plate type
  • Logging vehicles at entry and exit
Sizes
a YOLOv8 detector plus separate character-reading modules
Hardware
from: Laptop
Commercial use with conditionsDetails
Plates and vehicles2023–2024

ViT Vehicle Classifiers (dima806)

dima806 (open project) · not disclosed

Two simple photo classifiers: vehicle type and car brand. Trained on public datasets, so check them on your own footage before use. Filming vehicles and people is personal data processing: agree the procedure with a lawyer.

  • Identifying the vehicle type from a photo
  • Pre-labelling the car brand
  • Going through photos in listings and applications
Sizes
based on ViT-base (google/vit-base-patch16-224-in21k)
Hardware
from: Laptop
Commercial use allowedDetails

Translation 11

TranslationRUGGUF2026

TuvanGemma

Mergen Kungaa · not disclosed

Russian to Tuvan translation that runs on your own machine: ready GGUF builds, a LoRA adapter and a standalone Windows app. The author warns that names, numbers and important text should be double-checked with a native speaker.

  • Translating documents and notices into Tuvan
  • Translating Tuvan texts into Russian
  • Draft translation without sending data outside
Sizes
4B and 12B
Hardware
from: Laptop
Commercial use with conditionsDetails
TranslationRU2022–2026

SLONE: перевод для языков народов России

SLONE (David Dale, Aigiz Kunafin and others) · not disclosed

A community that extends NLLB and mBART to low-resource languages. Its models cover Erzya, Tuvan, Bashkir, Tatar, Chuvash, Buryat, Mari, Khakas and Karachay-Balkar. Quality differs by language, so test it on your own texts.

  • Translating from Russian into a national language and back
  • Preparing bilingual materials and signage
  • Aligning parallel texts
Sizes
42M for the compact encoders, 620M – 758M for the translators
Hardware
from: Laptop
Commercial use with conditionsDetails
TranslationRU2025–2026

Hunyuan-MT / Hy-MT

Tencent · China

Tencent translators for 33 languages; the first version won the WMT25 competition. Russian is supported. The small 1.8B version runs on a laptop; the new Hy-MT2 is under Apache 2.0.

  • Translating documents while keeping formatting
  • Translation with a set glossary of terms
  • Translating correspondence with Chinese partners
Sizes
1.8B – 30B-A3B
Hardware
from: Laptop
Commercial use with conditionsDetails
TranslationRU2020–2026

OPUS-MT (MarianMT)

Helsinki-NLP, University of Helsinki · Finland

More than a thousand small translators, each for its own language pair. Russian-English and back are available. Fast even on a regular CPU.

  • Bulk translation of short texts
  • Translation right on the server without a GPU
  • Translating reviews and requests before analysis
Sizes
25M – 240M
Hardware
from: Laptop
Commercial use allowedDetails
TranslationRUOllama2026

TranslateGemma

Google · USA

Translators based on Gemma 3 for 55 languages that can also translate text in images. Russian is supported. The 4B version fits on a laptop.

  • Translating documents and correspondence
  • Translating text from screenshots and photos
  • Localizing websites and apps
Sizes
4B – 27B
Hardware
from: Laptop
Commercial use with conditionsDetails
TranslationRUGGUF2025

Seed-X

ByteDance Seed · China

A compact ByteDance translator for 28 languages, close in quality to large closed systems. Russian is supported. Ready-made compressed versions are available.

  • Translating business correspondence and documents
  • Translating product cards
  • Translating technical and legal texts
Sizes
7B
Hardware
from: Laptop
Commercial use allowedDetails
TranslationRUGGUF2024–2025

Tower

Unbabel · Portugal

Language models tailored for translation and multilingual text work: translating, editing, and assessing translation quality. Russian is supported. Non-commercial license.

  • Translation that respects context and terminology
  • Post-editing machine translation
  • Assessing the quality of a finished translation
Sizes
2B – 72B
Hardware
from: Laptop
Non-commercial onlyDetails
TranslationRU2024

kazRush (deepvk)

deepvk (VK) · Russia

Compact Kazakh-Russian translators from VK. At 197M they translate as well as the 600M NLLB and run on a regular CPU.

  • Translating requests from Kazakh to Russian
  • Translating documents and instructions into Kazakh
  • Bilingual customer support
Sizes
197M
Hardware
from: Laptop
Commercial use allowedDetails
TranslationRUGGUFNot maintained2023

MADLAD-400

Google · USA

Google's translator for more than 400 languages under a permissive license. Russian is supported. A good substitute for NLLB when commercial use is needed.

  • Translating documents and emails
  • Translating catalogs and product descriptions
  • Translating into CIS and Asian languages
Sizes
3B – 10B
Hardware
from: Laptop
Commercial use allowedDetails
TranslationRUGGUFNot maintained2022–2023

NLLB-200

Meta · USA

A translator for 200 languages, including rare and minor ones. Russian is supported. Strong language coverage, but the license prohibits commercial use.

  • Translating texts between 200 languages
  • Translating into rare languages where no other models exist
  • Comparing quality when choosing a translator
Sizes
600M to 3.3B (plus 54B MoE)
Hardware
from: Laptop
Non-commercial onlyDetails
TranslationRUGGUFNot maintained2020–2021

M2M-100

Meta · USA

An early Meta translator that translates directly between 100 languages, without English in the middle. Russian is supported. Old, but light and permissively licensed.

  • Translation between any pair of 100 languages
  • Quick draft translation on modest hardware
  • Base for fine-tuning to your subject area
Sizes
418M – 12B
Hardware
from: Laptop
Commercial use allowedDetails

Moderation and safety 11

Moderation and safety2024–2026

GLiNER-PII

GLiNER community (Fastino, Knowledgator, NVIDIA and others) · USA

Small GLiNER-based models for finding personal data: passports, phone numbers, accounts, addresses. Data types are set in words. Russian is not officially supported.

  • Masking personal data before cloud AI
  • Finding passport data and bank details in documents
  • Checking data exports for leaks
Sizes
about 200M to 500M
Hardware
from: Laptop
Commercial use allowedDetails
Moderation and safety2024–2026

Aegis / Nemotron Safety Guard

NVIDIA · USA

NVIDIA content filters for bots, with separate models for keeping the conversation on topic and detecting jailbreaks. Safety Guard v3 was trained on 9 languages; Russian was tested only without fine-tuning.

  • Checking bot requests and replies
  • Keeping the bot within its topic
  • Detecting attempts to bypass rules
Sizes
4B – 8B
Hardware
from: Laptop
Commercial use with conditionsDetails
Moderation and safetyOllama2024–2026

Granite Guardian

IBM · USA

IBM judge models: they catch harm, profanity and jailbreak attempts, and in RAG and agents check whether an answer is grounded in the documents. You can state your own rule in words.

  • Checking bot requests and replies
  • Finding made-up facts in knowledge-base answers
  • Checking your own rules written as text
Sizes
38M – 8B
Hardware
from: Laptop
Commercial use allowedDetails
Moderation and safety2026

OpenAI Privacy Filter

OpenAI · USA

Finds and hides personal data: names, addresses, phone numbers, emails, account numbers, passwords. Runs even in the browser. Trained mostly on English.

  • Removing personal data from text before sending it to cloud AI
  • Finding passwords and keys in texts
  • Anonymizing correspondence for analytics
Sizes
1.5B (50M active)
Hardware
from: Laptop
Commercial use allowedDetails
Moderation and safety2025

AprielGuard

ServiceNow · USA

A guard model that catches both harmful content and attacks on AI (prompt injection, jailbreaks), including when agents use tools.

  • Screening chatbot requests for attacks and jailbreaks
  • Filtering harmful model answers
  • Monitoring the actions of AI agents that use tools
Sizes
8B
Hardware
from: Laptop
Commercial use allowedDetails
Moderation and safetyRUGGUF2025

Qwen3Guard

Alibaba (Qwen) · China

Safety filters for 119 languages, Russian among them. The Stream version checks a bot's reply while it is being generated and can cut it off on the fly.

  • Filtering bot requests in Russian
  • Stopping a dangerous reply during generation
  • Labeling messages by risk category
Sizes
0.6B – 8B
Hardware
from: Laptop
Commercial use allowedDetails
Moderation and safetyOllama2023–2025

Llama Guard

Meta · USA

Filter models that check chatbot requests and replies for dangerous topics against a list of categories. Version 4 also checks images. Russian is not officially supported.

  • Checking user questions to the bot
  • Checking bot replies before sending
  • Reporting which rule category was violated
Sizes
1B – 12B
Hardware
from: Laptop
Commercial use with conditionsDetails
Moderation and safety2023–2025

NSFW-классификаторы (Falconsai, Freepik)

Falconsai and Freepik · USA and Spain

Small models that tell explicit images from regular ones. The Freepik model distinguishes four levels of explicitness. They run on a CPU.

  • Filtering user photos and avatars
  • Checking generated images before publishing
  • Labeling a media library
Sizes
86M
Hardware
from: Laptop
Commercial use allowedDetails
Moderation and safety2024–2025

Prompt Guard

Meta · USA

Tiny classifiers that catch attempts to hack a bot: prompt injections and rule bypassing. The 86M version is multilingual, 22M is English only.

  • Protecting a bot from prompt injections
  • Checking emails and documents that reach an AI agent
  • Fast filter in front of a large model
Sizes
22M – 86M
Hardware
from: Laptop
Commercial use with conditionsDetails
Moderation and safetyOllama2024–2025

ShieldGemma

Google · USA

Gemma-based filters: they check text for dangerous and offensive content, and ShieldGemma 2 checks images. Focused on English.

  • Moderating user messages
  • Checking bot replies
  • Checking generated images before publishing
Sizes
2B – 27B
Hardware
from: Laptop
Commercial use with conditionsDetails
Moderation and safety2024

Piiranha

iiiorg · not disclosed

A popular detector of 17 types of personal data in six European languages. No Russian and a non-commercial license: suitable for trials and research.

  • Finding personal data in texts
  • Comparing the quality of PII detectors
Sizes
278M
Hardware
from: Laptop
Non-commercial onlyDetails

Tabular data 11

Tabular data2026

Kumo Tabular

NVIDIA · USA

A pretrained model for tables: classification and regression from in-context examples, with no task-specific training. You supply labeled rows and the rows that need a forecast and get class probabilities.

  • scoring customers and requests
  • forecasting from a table of sales or orders
  • detecting churn from a customer table
Sizes
size not stated on the card
Hardware
from: Laptop
Commercial use allowedDetails
Tabular data2026

Xiaomi-TabLDM

Xiaomi · China

A small tabular model: it takes labeled rows as context and predicts values for new rows in one pass, with no task-specific training. Two checkpoints: classification and regression.

  • forecasts from a table of customers or orders
  • estimating price or demand from sales rows
  • scoring requests without training a separate model
Sizes
about 70M
Hardware
from: Laptop
Commercial use with conditionsDetails
Tabular data2025–2026

LimiX

Stable AI (Beijing, with Tsinghua University) · China

A table model that alone can classify, predict numbers and fill in missing data. The lightweight LimiX-2M runs on an ordinary computer.

  • Filling gaps in 1C and CRM exports
  • Churn prediction and scoring
  • Classifying customers and products
Sizes
2M – 16M and LimiX-2
Hardware
from: Laptop
Commercial use with conditionsDetails
Tabular data2025–2026

Mitra

Amazon (AutoGluon team) · USA

Amazon's tabular model built into AutoGluon: classification and regression from examples with brief fine-tuning. Mitra-v2 handles more rows and columns.

  • Predicting churn and repeat purchases
  • Scoring applications
  • Predicting deal or order value
Sizes
about 76M
Hardware
from: Laptop
Commercial use allowedDetails
Tabular data2025–2026

TabDPT

Layer 6 AI (TD Bank) · Canada

A tabular model from a Canadian bank's AI lab, trained on real tables rather than only synthetic ones. Version 1.2 Turbo made computation orders of magnitude faster.

  • Scoring applications and customers
  • Predicting churn
  • Classifying transactions and customers
Sizes
about 60–80M
Hardware
from: Laptop
Commercial use allowedDetails
Tabular data2022–2026

TabPFN

Prior Labs (University of Freiburg) · Germany

A ready-made model for tables: it takes example rows and immediately predicts for new ones, without lengthy training or tuning. Only v2 is free for business; newer versions are non-commercial.

  • Predicting customer churn from a CRM export
  • Scoring applications and leads
  • Classifying customers from 1C data
Sizes
from a few to hundreds of millions of parameters
Hardware
from: Laptop
Commercial use with conditionsDetails
Tabular data2026

EXAONE Tabular

LG AI Research · South Korea

LG's small tabular model: with 21M parameters it nearly matches the leaders in classification and regression accuracy. Weights are for non-commercial use only.

  • Pilot churn forecasts
  • Testing scoring hypotheses
  • Exploring customer data
Sizes
about 21M
Hardware
from: Laptop
Non-commercial onlyDetails
Tabular data2025–2026

Orion-MSP / Orion-BiX

Lexsi Labs · India

Recent open models for tabular data: they predict from a few examples given in the prompt, with no task-specific training.

  • Classification and forecasting on tables with no separate training
  • Quickly testing models on new datasets
  • Assessing features in large tables
Sizes
size not stated on the model card
Hardware
from: Laptop
Commercial use allowedDetails
Tabular data2025–2026

TabICL

Inria (Soda team) · France

An open tabular model from the creators of scikit-learn: classifies and predicts from examples without training and handles tables of up to hundreds of thousands of rows. The license allows business use.

  • Predicting customer churn
  • Scoring applications and deals
  • Classifying customers from 1C and CRM data
Sizes
about 25–30M
Hardware
from: Laptop
Commercial use allowedDetails
Tabular dataGGUF2024

TableLLM

RUCKBReasoning, Renmin University of China · China

A model for office work with tables: for a given question it returns either a direct answer or code to process the data in a table or document.

  • Processing tables from Excel and documents from a text instruction
  • Generating code for recalculations and selections
  • Answering questions about data in reports
Sizes
7B и 13B
Hardware
from: Laptop
Commercial use with conditionsDetails
Tabular dataNot maintained2023

TableLlama

OSU NLP Group, Ohio State University · USA

A general-purpose model for tables: filling gaps, finding rows, matching columns and answering questions about the data.

  • Answering questions about tables inside documents
  • Matching columns across different tables
  • Finding and completing records in reference books
Sizes
7B
Hardware
from: Laptop
Commercial use with conditionsDetails

Text to SQL 11

Text to SQLGGUF2024–2025

Prem-1B-SQL

Prem AI · UK

A text-to-SQL model of just 1B parameters, designed to run locally so the database never leaves for external services.

  • Local translation of questions into SQL with no internet access
  • Query hints on modest hardware
  • Embedding into internal analytics tools
Sizes
1B
Hardware
from: Laptop
Commercial use allowedDetails
Text to SQLGGUF2025

SQL-R1

IDEA Research · China

A text-to-SQL model trained with reinforcement learning: it works through the schema and the conditions step by step before producing a query.

  • Database queries for questions with several conditions
  • Reviewing and fixing other people SQL queries
  • An analyst helper inside a BI system
Sizes
3B – 14B
Hardware
from: Laptop
Commercial use allowedDetails
Text to SQL2025

Arctic-Text2SQL-R1

Snowflake · USA

A Snowflake model for turning questions into SQL, trained with reinforcement learning by checking query results. The open 7B version is based on Qwen2.5-Coder.

  • Plain-language questions to a data warehouse
  • Generating SQL for reports and dashboards
  • Checking and fixing analysts' queries
Sizes
7B
Hardware
from: Laptop
Commercial use allowedDetails
Text to SQL2025

XiYanSQL-QwenCoder

Alibaba · China

Alibaba models for turning questions into SQL, based on Qwen2.5-Coder. They work with different SQL dialects; a small 3B version suits modest hardware.

  • Plain-language database questions
  • Queries for different databases (PostgreSQL, MySQL, SQLite)
  • Automating routine reports
Sizes
3B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Text to SQL2025

OmniSQL

Renmin University of China (RUC) · China

Models for turning questions into SQL, trained on millions of synthetic query examples across different databases. Three sizes for different hardware.

  • Database questions without knowing SQL
  • Generating queries for reports
  • A base for fine-tuning on your own database schema
Sizes
7B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Text to SQLOllamaNot maintained2023–2024

SQLCoder

Defog · USA

One of the first open models that turn a plain-language question into an SQL query against a database. Available in Ollama, but newer competitors are already stronger.

  • Answering managers' questions from the sales database without an analyst
  • Drafting SQL queries for reports
  • An assistant inside a BI system
Sizes
7B – 70B
Hardware
from: Laptop
Commercial use with conditionsDetails
Text to SQLGGUFNot maintained2024

Chat2DB-SQL

Chat2DB · China

A text-to-SQL model from the open Chat2DB database client: it supports different SQL dialects, with an English and Chinese model card.

  • Turning a question into SQL inside a database client
  • Drafting queries for different database engines
  • Hints for developers working with a schema
Sizes
7B
Hardware
from: Laptop
Commercial use allowedDetails
Text to SQLGGUFNot maintained2024

Natural-SQL

ChatDB · USA

A text-to-SQL model built on DeepSeek-Coder, aimed at complex questions spanning several tables and conditions.

  • Complex queries joining several tables
  • Answering database questions without an analyst
  • Drafting SQL for reports and exports
Sizes
7B
Hardware
from: Laptop
Commercial use with conditionsDetails
Text to SQLOllamaNot maintained2024

DuckDB-NSQL

MotherDuck and Numbers Station · USA

A model for turning questions into SQL, built for the embedded analytics database DuckDB. Available in Ollama, convenient for working with CSV and Parquet locally.

  • Plain-language questions about CSV and Parquet exports
  • DuckDB queries inside analytics scripts
  • Quick analytics on a laptop without a server
Sizes
7B
Hardware
from: Laptop
Commercial use with conditionsDetails
Text to SQLNot maintained2023

CodeS

RUCKBReasoning, Renmin University of China · China

An early line of open text-to-SQL models starting at 1B, including variants fine-tuned for specific database schemas.

  • Turning an employee question into an SQL query
  • Drafting warehouse queries for a report
  • Embedding into a BI dashboard as a helper
Sizes
1B – 15B
Hardware
from: Laptop
Commercial use allowedDetails
Text to SQLGGUFNot maintained2023

NSQL

Numbers Station · USA

One of the first open text-to-SQL lines, including very small versions from 350M that run on an ordinary PC.

  • Turning a question into SQL from a table description
  • Hints while writing queries
  • A local analyst helper with no data leaving the company
Sizes
350M – 7B
Hardware
from: Laptop
Commercial use with conditionsDetails

Music and sound 9

Music and soundGGUF2022–2026

MERT

m-a-p (Multimodal Art Projection) · UK / China

A music encoder: turns a track into a numeric representation used to detect genre, mood, key and rhythm. MERT-v2 handles full songs up to 6 minutes.

  • Automatic tagging of a music catalog
  • Finding similar tracks
  • Detecting genre, mood and tempo
Sizes
95M – 632M
Hardware
from: Laptop
Non-commercial onlyDetails
Music and soundRUGGUF2025–2026

ACE-Step

ACE Studio and StepFun · China

Fast generation of songs with vocals in 19 languages, including Russian: a full song in seconds, editing of individual parts and style changes.

  • Songs and jingles for ads
  • Background music for videos
  • Demo versions of tracks
Sizes
about 2B – 4B
Hardware
from: Laptop
Commercial use allowedDetails
Music and soundGGUF2024–2026

Stable Audio

Stability AI · UK

Generates short music clips and sound effects from a description. Version 3 is split into separate models for music and for sounds.

  • Sound effects for videos and games
  • Background music and jingles
  • Interface sounds
Sizes
about 0.5B – 2.3B
Hardware
from: Laptop
Commercial use with conditionsDetails
Music and sound2025

DiffRhythm

ASLP-lab (Northwestern Polytechnical University) · China

Fast generation of a full song with vocals from lyrics and a style sample, up to several minutes long.

  • Songs and jingles from lyrics
  • Music for videos
  • Demo versions of tracks
Sizes
about 1.1B
Hardware
from: Laptop
Commercial use allowedDetails
Music and sound2024

MMAudio

University of Illinois and Sony AI · USA / Japan

Adds sound to silent video: generates noises and sound effects in sync with the on-screen action, from the video and a text prompt. One of the first strong open Foley models.

  • Sound effects for silent video
  • Sound for clips from AI generators
  • Draft sound design for editing
Sizes
size not stated on the model card
Hardware
from: Laptop
Non-commercial onlyDetails
Music and sound2024

MuQ / MuQ-MuLan

Tencent AI Lab · China

The MuQ music encoder and the MuQ-MuLan model, which matches music and text: you can search for tracks by a description in English or Chinese.

  • Searching music by text description
  • Tagging tracks by genre and mood
  • Finding similar music
Sizes
300M – 700M
Hardware
from: Laptop
Non-commercial onlyDetails
Music and sound2023–2024

MusicGen (AudioCraft)

Meta · USA

Generates instrumental music from a text description or a sample melody. One of the first open models of its kind.

  • Draft music sketches
  • Music for video prototypes
  • Research
Sizes
300M – 3.3B
Hardware
from: Laptop
Non-commercial onlyDetails
Music and soundNot maintained2023

CLAP (LAION)

LAION · Germany

CLIP for audio: maps audio and text into a shared space. Lets you search sounds and music by description and classify them without training. Text must be in English.

  • Search sounds and music by description
  • Automatic tags for an audio library
  • Recognizing sound types (siren, breaking glass, voice)
Sizes
size not stated on the model card
Hardware
from: Laptop
Commercial use allowedDetails
Music and soundNot maintained2022

AST (Audio Spectrogram Transformer)

MIT · USA

A classic 2021 sound recognition model: detects 527 AudioSet event classes (siren, barking, breaking glass, music). Lightweight, runs without a GPU, in Transformers since 2022.

  • Sound event recognition
  • Tagging an audio archive
  • Detecting alarm sounds
Sizes
about 87M
Hardware
from: Laptop
Commercial use allowedDetails

Computer-use agents 9

Computer-use agentsGGUF2025–2026

UI-Venus

Ant Group (inclusionAI) · China

An Ant Group family for finding elements on screen and completing tasks in phone and computer interfaces. UI-Venus-2 was specifically trained to refuse dangerous actions. In October LBS-27B came out (Apache 2.0, based on Qwen3.8-27B): it learns to pick and zoom into the needed parts of an image on interface markup, and carries this skill over to ordinary questions about pictures.

  • Automating actions in mobile apps
  • Filling in forms in web interfaces
  • UI autotests
Sizes
2B – 72B
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer-use agentsGGUF2025–2026

Fara

Microsoft · USA

Small Microsoft models for working in the browser: they look at the page and click, type and scroll. Designed to run directly on a work computer without the cloud.

  • Filling in web forms and applications
  • Collecting data from web portals without an API
  • Checking websites against scenarios
Sizes
4B – 27B
Hardware
from: Laptop
Commercial use allowedDetails
Computer-use agentsGGUF2025–2026

Holo

H Company · France

A French model family for controlling a browser and computer: precisely finds the right element on screen and handles multi-step tasks. The latest Holo3 and 3.1 are open under Apache 2.0.

  • Working in web portals and legacy software without an API
  • Filling in forms and applications
  • Testing interfaces against scenarios
Sizes
0.8B – 235B-A22B
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer-use agents2024–2026

ShowUI

Show Lab (National University of Singapore) · Singapore

A lightweight model for working with interfaces: finds buttons and fields by description and performs actions on the web and on a phone. ShowUI-π can drag with the mouse.

  • Clicking and filling in forms from a task description
  • Web UI autotests
  • An assistant on a low-end computer without the cloud
Sizes
2B (ShowUI), about 500M (ShowUI-π)
Hardware
from: Laptop
Commercial use with conditionsDetails
Computer-use agents2025–2026

GUI-Owl / Mobile-Agent

Alibaba (Tongyi Lab, X-PLUG) · China

Models for controlling phones and computers from the Mobile-Agent project: they work with Android, Windows, macOS and the browser; version 1.5 has a reasoning mode.

  • Automating actions in mobile apps
  • Working in desktop software without an API
  • Testing apps against scenarios
Sizes
2B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Computer-use agentsGGUF2025

MAI-UI

Alibaba (Tongyi-MAI) · China

Compact Alibaba models for working in smartphone and computer interfaces: they find elements and complete multi-step tasks. The agent acts on the user's behalf inside their apps, so access rights and limits are set before launch. The follow-up line, Qwen-UI-Agent, was announced in July 2026, but its weights have not been released.

  • Automating actions in mobile apps
  • Working in software without an API
  • UI autotests
Sizes
2B – 8B
Hardware
from: Laptop
Commercial use allowedDetails
Computer-use agentsGGUF2025

UI-TARS

ByteDance Seed · China

A model that looks at a screenshot and controls the mouse and keyboard itself: clicks, fills in fields, navigates menus. The first generation and 1.5-7B are open; UI-TARS-2 weights were not released.

  • Working in legacy software without an API
  • Filling in forms and moving data between systems
  • UI autotests from plain-language scenarios
Sizes
2B – 72B
Hardware
from: Laptop
Commercial use allowedDetails
Computer-use agents2024–2025

xLAM

Salesforce · USA

Salesforce models for function calling and agents: they pick the right tool and fill in its parameters. Strong on benchmarks, but the license is non-commercial.

  • Calling APIs and internal services on user request
  • Multi-step agents with several tools
  • Comparing approaches before choosing a commercial model
Sizes
1B – 8x22B
Hardware
from: Laptop
Non-commercial onlyDetails
Computer-use agents2024–2025

OmniParser

Microsoft · USA

Breaks a screenshot down into buttons, fields and icons with labels so a regular language model can understand and control the screen. It does not click itself; it serves as the agent's eyes.

  • Mapping legacy software screens for automation
  • Preparing an agent to work in an interface
  • Checking that the required elements are on screen
Sizes
under 1B (detector + captioning)
Hardware
from: Laptop
Commercial use with conditionsDetails

Cybersecurity 9

Cybersecurity2025–2026

Foundation-Sec

Cisco (Foundation AI) · USA

Cisco models for information security based on Llama 3.1 8B: analysis of vulnerabilities, threats and incidents. Can be deployed inside your own perimeter.

  • Analyzing vulnerability and threat reports
  • Helping SOC analysts during incidents
  • Mapping threats to MITRE ATT&CK
Sizes
8B
Hardware
from: Laptop
Commercial use with conditionsDetails
CybersecurityGGUF2023–2025

SecGPT

Clouditera · China

A Chinese open family for cybersecurity: reviewing vulnerabilities, analysing logs and traffic, explaining commands and scripts.

  • Reviewing vulnerabilities and drafting fix recommendations
  • Analysing logs and reconstructing an attack chain
  • Explaining suspicious commands and scripts
Sizes
1.5B – 14B
Hardware
from: Laptop
Commercial use allowedDetails
CybersecurityGGUF2023–2025

WhiteRabbitNeo / DeepHat

Kindo · USA

One of the best-known open families for security and DevSecOps work: reviewing code for weaknesses, test scenarios, explaining attacks.

  • Finding weak spots in code and configurations
  • Reviewing incidents and explaining attack techniques
  • Drafting scripts and procedures for the security team
Sizes
7B – 70B
Hardware
from: Laptop
Commercial use with conditionsDetails
Cybersecurity2025

Llama-Primus

Trend Micro · Japan

Trend Micro cybersecurity models based on Llama 3.1 8B, fine-tuned on a corpus of security texts. A reasoning version is available.

  • Answering questions about threats and vulnerabilities
  • Analyzing cyberattack reports
  • A base for fine-tuning for SOC tasks
Sizes
8B
Hardware
from: Laptop
Commercial use with conditionsDetails
Cybersecurity2024

Phishing Email Detection DistilBERT

cybersectony · not disclosed

A very light classifier for emails and links showing signs of phishing. It errs in both directions, so borderline emails are still reviewed by a person.

  • Flagging suspicious incoming emails
  • Checking links from correspondence before opening them
  • A first-level filter in a mail gateway
Sizes
about 66M
Hardware
from: Laptop
Commercial use allowedDetails
Cybersecurity2023–2024

Biglog

Huawei and University of Science and Technology of China (LogAIBox project) · China

A model pretrained on a large volume of logs from different systems: it turns a log line into a vector on top of which failure and outlier detection is fine-tuned. A human makes the decision on a disputed operation; automatic blocking without review is not acceptable.

  • Finding unusual entries in operation logs
  • Grouping similar failure messages
  • Preparing features for an in-house monitoring model
Sizes
about 110M (BERT-base architecture)
Hardware
from: Laptop
Commercial use with conditionsDetails
CybersecurityNot maintained2023–2024

LogPrecis / SecureShellBert

SmartData@PoliTO, Politecnico di Torino · Italy

A pair of models from an Italian university: they read a recorded command session and mark which parts belong to reconnaissance, persistence and other attacker steps. A human makes the decision on a disputed operation; automatic blocking without review is not acceptable.

  • Labelling command logs by stage of activity
  • Grouping similar sessions for incident review
  • Preparing features for server monitoring
Sizes
about 125M (based on CodeBERT)
Hardware
from: Laptop
Commercial use with conditionsDetails
CybersecurityGGUFNot maintained2023–2024

ZySec

ZySec AI · India

A small open assistant for security professionals: questions about standards, reviewing threats and vulnerabilities, drafting internal documents.

  • Answering questions about security policies and standards
  • First-pass review of threat reports
  • Drafting internal protection guidelines
Sizes
2.8B и 7B
Hardware
from: Laptop
Commercial use allowedDetails
CybersecurityNot maintained2022–2023

SecureBERT

Ehsan Aghaei and co-authors · USA

A compact language encoder trained on cybersecurity texts: tagging threat reports, finding entities and classification.

  • Tagging threat reports and vulnerability bulletins
  • Extracting entities from security texts
  • Classifying and searching an internal incident base
Sizes
около 125M
Hardware
from: Laptop
Commercial use with conditionsDetails

Recommenders 9

RecommendersGGUF2025–2026

OneRec (OpenOneRec)

Kuaishou · China

An open foundation model for recommendations from a short-video service: browsing and purchase history is fed to the model as a sequence and it predicts the next item or clip. Fine-tuning on your own data is required.

  • A personal feed of products or content
  • Picking the next item from order history
  • Recommendations inside ad slots
Sizes
1.7B, 8B
Hardware
from: Laptop
Commercial use allowedDetails
Recommenders2025

MiniOneRec

MiniOneRec open project · China

A small open generative recommender at 1.5B with published checkpoints and a full training pipeline. Handy as a learning and starter option; a real store needs fine-tuning on its own orders.

  • A quick recommendation prototype on your own data
  • Testing ideas before running a large model
  • Training a team to work with generative recommendations
Sizes
1.5B
Hardware
from: Laptop
Commercial use allowedDetails
Recommenders2024

Marqo Ecommerce Embeddings

Marqo · Australia

Models that turn a product photo and its title into a single vector: suitable for a similar-items block, image search and matching by description. Built for retail product cards.

  • Similar items by photo and title
  • Catalogue search by image
  • Product search from a shopper phrase
Sizes
203M (B) и 652M (L)
Hardware
from: Laptop
Commercial use allowedDetails
Recommenders2024

beeFormer

Recombee · Czechia

Description encoders trained so that items people actually buy together end up close to each other, not merely items that sound alike. Versions trained on public catalogues are published.

  • A frequently-bought-together block
  • Recommendations for listings with no sales history
  • Moving the model to a new catalogue
Sizes
about 110M (built on all-mpnet-base-v2)
Hardware
from: Laptop
Commercial use with conditionsDetails
RecommendersNot maintained2024

EasyRec (HKUDS)

HKUDS, University of Hong Kong · China

Encoders that put a text portrait of a shopper and a text description of an item into one vector space so they can be matched directly. They help where a new listing has no purchase statistics yet.

  • Recommendations for new items with no statistics
  • Matching by a text portrait of the shopper
  • Reinforcing ordinary recommendations with descriptions
Sizes
three RoBERTa-based versions: small, base, large
Hardware
from: Laptop
Commercial use allowedDetails
RecommendersGGUFNot maintained2024

eCeLLM

Ohio State University (NingLab) · USA

Three language models tuned on a large set of commerce instructions: product selection, answers from the item card, matching listings, review analysis. Your own assortment still needs fine-tuning.

  • Finding a product from a shopper request
  • Answering questions about an item card
  • Matching identical listings across catalogues
Sizes
2.7B – 13B
Hardware
from: Laptop
Commercial use with conditionsDetails
RecommendersNot maintained2024

BLaIR

UC San Diego (McAuley Lab) · USA

Light encoders trained on pairs of item card and real review: they produce item vectors for similar-item blocks and for search by a shopper description. The weights are small and run on an ordinary server.

  • A similar-items block based on descriptions
  • Product search from a shopper phrase
  • Suggesting cards for new listings
Sizes
125M и 355M
Hardware
from: Laptop
Commercial use allowedDetails
RecommendersNot maintained2023

FashionCLIP

FashionCLIP research project (Farfetch data) · not disclosed

A time-tested model for clothing and footwear: it places an item photo and its description in one space, so it fits similar-item blocks and outfit matching. Fine-tuning for your own assortment is advisable.

  • Similar items by photograph
  • Building an outfit and matching add-ons
  • Automatic tagging of clothing cards
Sizes
built on CLIP ViT-B/32
Hardware
from: Laptop
Commercial use allowedDetails
RecommendersNot maintained2022

P5

Rutgers University · USA

An early and still popular approach where recommendations are expressed as plain text: one small engine handles selection, rating and explanation. Checkpoints are trained on public datasets; a store needs its own.

  • Suggesting the next item from history
  • Predicting an item rating
  • A text explanation of a recommendation
Sizes
small and base versions built on T5
Hardware
from: Laptop
Commercial use allowedDetails

Fact-checking and judges 8

Fact-checking and judgesRU2025–2026

POLLUX Judge (ai-forever)

SberDevices (ai-forever) · Russia

Judge models that evaluate other AI models' answers in Russian: they score against a given criterion and explain the score in text.

  • Automatic quality checks of Russian chatbot answers
  • Comparing several models before choosing one
  • Checking answers after fine-tuning
Sizes
4B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Fact-checking and judges2024–2025

CompassJudger / CompassVerifier

OpenCompass (Shanghai AI Laboratory) · China

A line of judges from the team behind open model benchmarks: they score answers and check them against a reference. The judge itself makes mistakes and does not replace manual review on important tasks.

  • Scoring model answers against set criteria
  • Checking an answer against a reference solution
  • Comparing several models on your own data
Sizes
1.5B – 32B
Hardware
from: Laptop
Commercial use allowedDetails
Fact-checking and judges2024–2025

Skywork-Reward

Skywork (Kunlun Tech) · China

Reward models: they score how good a language model's answer is for the user. Used for fine-tuning your own models and picking the best of several answers.

  • Choosing the best of several bot answers
  • Scoring answer quality during model fine-tuning
  • Comparing models before rollout
Sizes
0.6B – 27B
Hardware
from: Laptop
Commercial use with conditionsDetails
Fact-checking and judges2024

GLIDER (Patronus)

Patronus AI · USA

A small judge: it scores against your criteria and highlights which part of the answer led to that score. The license is non-commercial. The judge itself makes mistakes and does not replace manual review.

  • Scoring answers against your criteria with an explanation
  • Understanding why a score was lowered
  • Bulk review of assistant conversations
Sizes
3.8B (based on Phi-3.5-mini)
Hardware
from: Laptop
Non-commercial onlyDetails
Fact-checking and judges2024

Flow Judge

Flow AI · not disclosed

A small judge model: it checks an answer against your instruction and gives a score with an explanation. Fits on a modest server. The judge itself makes mistakes and does not replace manual review.

  • Checking AI assistant answers against the instruction
  • Bulk scoring of exported conversations
  • Quality control before rolling out changes
Sizes
3.8B (based on Phi-3.5-mini)
Hardware
from: Laptop
Commercial use allowedDetails
Fact-checking and judgesOllamaNot maintained2024

MiniCheck

UT Austin and Bespoke Labs · USA

Checks whether each claim in an AI answer is supported by the source documents. The small versions are free; the larger 7B is in Ollama but non-commercial.

  • Checking RAG bot answers against documents
  • Finding unsupported claims in reports and summaries
  • Automated quality control of AI answers
Sizes
0.4B – 7B
Hardware
from: Laptop
Commercial use with conditionsDetails
Fact-checking and judgesNot maintained2023–2024

HHEM (Vectara)

Vectara · USA

A small model that checks whether an AI answer is grounded in the source text or made up. Runs on a CPU and works well as a filter in RAG systems.

  • Checking knowledge base chatbot answers for fabrications
  • Quality control of document summaries
  • Comparing language models by their tendency to make errors
Sizes
110M
Hardware
from: Laptop
Commercial use allowedDetails
Fact-checking and judgesNot maintained2023–2024

Prometheus 2

KAIST and LG AI Research (prometheus-eval) · South Korea

An open judge model: it scores other models' answers against your criteria and explains the score. A replacement for paid models in the reviewer role.

  • Scoring chatbot answers on your own scale
  • Comparing two answer options
  • Quality checks before launching an AI service
Sizes
7B – 8x7B
Hardware
from: Laptop
Commercial use allowedDetails

Rerankers 8

RerankersRU2025–2026

Llama Nemotron Rerank

NVIDIA · USA

A small 1B reranker from NVIDIA. The vl version also takes document pages as images, not just text. The new 8B (Nemotron-3.5-Rerank) was evaluated on 34 languages, including Russian, and accepts text, page images or both, including charts and tables.

  • Reordering passages before an AI assistant answers
  • Sorting retrieved scan and PDF pages
  • Search across internal policies and instructions
Sizes
1B and 8B
Hardware
from: Laptop
Commercial use with conditionsDetails
RerankersGGUF2024–2026

Jina Reranker

Jina AI · Germany

Strong multilingual rerankers; m0 also ranks pages as images (scans, slides). The latest versions are open for non-commercial use only.

  • Refining search results before a chatbot answers
  • Sorting retrieved PDF pages and slides
  • Catalog and knowledge base search
Sizes
33M – 2.4B
Hardware
from: Laptop
Commercial use with conditionsDetails
Rerankers2026

R3 (R3-Embedding и R3-Rerank)

Tencent · China

A pair of small Tencent models based on Qwen3 that pick the right skill for an AI agent for a given request: the embedding model finds candidates, the reranker chooses the best one.

  • Choosing a tool or skill for an AI agent
  • Routing requests between bot scenarios
  • Search across a catalog of internal tools
Sizes
0.6B
Hardware
from: Laptop
Commercial use allowedDetails
Rerankers2025

zerank (ZeroEntropy)

ZeroEntropy · USA

Rerankers built on Qwen3. The model card lists the target domains — finance, law, code, medicine, science; the stated language is English.

  • Refining results before an AI assistant answers
  • Sorting search results across contracts and reports
  • Search across technical and scientific documentation
Sizes
zerank-2 — 4B (based on Qwen3-4B), plus a smaller "small" version
Hardware
from: Laptop
Commercial use allowedDetails
Rerankers2023–2025

ColBERT (поиск с поздним взаимодействием)

Stanford NLP, later Answer.AI and LightOn · USA and France

A different search principle: every word of the question is compared with every word of the document, not the two texts as a whole. The index is heavier than with ordinary embeddings. The model cards list English.

  • Search across a knowledge base of long documents
  • Reordering retrieved passages
  • Search across policies and technical documentation
Sizes
about 33M – 150M
Hardware
from: Laptop
Commercial use with conditionsDetails
RerankersGGUFNot maintained2023–2024

BGE Reranker

BAAI (Beijing Academy of Artificial Intelligence) · China

Rerankers: they take passages found by search and reorder them by how well they actually match the question. v2-m3 is multilingual and lightweight, often paired with bge-m3.

  • Refining search results before a chatbot answers
  • Sorting knowledge base search results
  • Selecting the most relevant clauses of contracts and policies
Sizes
278M – 9B
Hardware
from: Laptop
Commercial use allowedDetails
RerankersNot maintained2023

BCEmbedding (Youdao)

NetEase Youdao · China

An embedding-plus-reranker pair for knowledge bases. The card lists English, Chinese, Japanese and Korean — Russian is not among the stated languages.

  • Search across a knowledge base and reference materials
  • Reordering retrieved passages
  • Picking answers for a support chatbot
Sizes
about 280M
Hardware
from: Laptop
Commercial use allowedDetails
RerankersRUNot maintained2022

Cross-Encoder MS MARCO (MiniLM, TinyBERT, mMARCO)

UKP Lab and the Sentence Transformers community · Germany

The most downloaded open rerankers: a tiny model reads a question-passage pair and scores how well they match. The multilingual mMARCO version covers Russian.

  • Reordering knowledge base search results
  • Selecting passages before a chatbot answers
  • Finding duplicates among tickets and product cards
Sizes
about 4M – 120M
Hardware
from: Laptop
Commercial use allowedDetails

Drawings and CAD 8

Drawings and CADRUGGUF2026

GreenMap Blueprint

GreenMap · not disclosed

A set of small models for Russian-language construction drawings: it finds the drawing area, title block and legend, marks up walls and hatching, and pulls sheet data into JSON along with references to standards. A draft that an engineer checks.

  • Finding the title block, legend and drawing area on a sheet
  • Marking up walls and hatching on a floor plan
  • Extracting drawing data into structured JSON
Sizes
3.5M - 4B
Hardware
from: Laptop
Commercial use allowedDetails
Drawings and CAD2026

CADReasoner

FusionBrain Lab, Lomonosov Moscow State University and Innopolis University · Russia

It writes CadQuery code from images and a point cloud, then edits its own output over several rounds, comparing the resulting shape with the original. This is a draft model: an engineer checks and signs off the document.

  • Recovering a parametric model from photos and a scan of a part
  • Step-by-step refinement of a draft CAD model
  • Digitising parts that have no drawings
Sizes
2.2B
Hardware
from: Laptop
Commercial use allowedDetails
Drawings and CADGGUF2026

Zero-to-CAD

Autodesk Research · USA

A model from Autodesk Research: from eight rendered views of a part it writes an executable CadQuery program. It was trained on synthetic data only. It produces a draft; an engineer checks and signs off the document.

  • Getting a parametric model from a set of views of a part
  • Converting finished geometry into editable code
  • A draft model for a design office
Sizes
2B (based on Qwen3-VL-2B)
Hardware
from: Laptop
Commercial use allowedDetails
Drawings and CAD2025

SpatialLM

Manycore Research · China

From a point cloud of a room or a walkthrough video it builds a structured plan: walls, doors, windows and objects with coordinates and sizes. It is a draft survey; final areas are confirmed by a specialist.

  • Building a floor plan from a 3D scan of a space
  • Extracting walls, doors and windows with dimensions
  • Draft measurements before a renovation or a move
Sizes
0.5B - 1B
Hardware
from: Laptop
Commercial use with conditionsDetails
Drawings and CAD2025

cadrille

AIRI, Lomonosov Moscow State University and Innopolis University · Russia

A successor to CAD-Recode: it builds a CAD model as CadQuery code from three kinds of input at once - a point cloud, an image and a text description. It produces a draft; an engineer checks and signs off the document.

  • Building a CAD model from a photo of a part
  • Recovering a model from a 3D scan
  • A draft part from a short text description
Sizes
2.2B (based on Qwen2-VL)
Hardware
from: Laptop
Non-commercial onlyDetails
Drawings and CAD2024–2025

CAD-Recode

University of Luxembourg (SnT) and Artec3D · Luxembourg

Turns a point cloud of a part into executable CadQuery Python code, that is, a parametric 3D model you can open and edit in CAD. The output is a draft: an engineer checks and signs off the document.

  • Recovering a parametric model from a 3D scan of a part
  • Converting a point cloud into editable CadQuery code
  • Preparing a draft model for a design engineer to finish
Sizes
1.5B (Qwen2-1.5B plus a single projection layer)
Hardware
from: Laptop
Non-commercial onlyDetails
Drawings and CADGGUF2025

StarVector

ServiceNow Research, Mila and ETS Montreal · Canada

It turns a raster image into SVG code: instead of tracing pixels the model writes vector commands, so lines and symbols stay editable. It simplifies complex drawings, so the result needs a visual check.

  • Vectorising diagrams, symbols and logos from raster images
  • Turning scans of simple drawings into editable vectors
  • Recovering icons and pictograms from images
Sizes
1B and 8B
Hardware
from: Laptop
Commercial use allowedDetails
Drawings and CADNot maintained2023

MatCha и DePlot

Google Research · USA

Small models that turn an image of a chart or diagram into a table of numbers and answer questions about it. Values read off an image are approximate, so the result has to be checked.

  • Turning a chart from a report or PDF into a table
  • Natural-language questions about a diagram
  • A short text description of a chart
Sizes
282M
Hardware
from: Laptop
Commercial use allowedDetails

Finance 7

Finance2024–2026

AnomalyGFM

MaLA Lab, Singapore Management University (with University of Technology Sydney) · Singapore

A pretrained model that finds anomalous nodes in a relationship graph and marks suspicious vertices on a new graph without training on it. A human makes the decision on a disputed operation; automatic blocking without review is not acceptable.

  • Finding suspicious accounts and counterparties linked to each other
  • Highlighting atypical participants in a payment graph
  • Testing hypotheses about ring schemes and chains of intermediaries
Sizes
weights file about 1.5 MB
Hardware
from: Laptop
Commercial use with conditionsDetails
Finance2025

Kronos

Tsinghua University (NeoQuasar) · China

A foundation model for market candlestick data: trained on data from more than 45 exchanges, it forecasts prices and volumes. The largest version, large, is not open.

  • Forecasting candlesticks and trading volumes
  • Volatility estimation
  • A base for fine-tuning on your own series
Sizes
4.1M – 102M
Hardware
from: Laptop
Commercial use allowedDetails
Finance2025

Fin-R1

Shanghai University of Finance and Economics (SUFE) · China

A reasoning model for financial tasks based on Qwen2.5-7B: calculations, report analysis, regulatory questions. Trained on Chinese and English data.

  • Financial calculations with step-by-step explanations
  • Answering questions about financial statements
  • Analyzing tables of financial data
Sizes
7B
Hardware
from: Laptop
Commercial use allowedDetails
Finance2023–2024

FinMA (PIXIU)

The Fin AI / ChanceFocus · international project

One of the first open model families for financial text: reading statements, news and questions about numbers. Not investment advice: decisions are made by a specialist.

  • Reading financial statements and press releases
  • Answering questions about numeric data in documents
  • Classifying financial texts
Sizes
0.5B – 30B
Hardware
from: Laptop
Commercial use with conditionsDetails
Finance2024

RevTrack / RevClassify

MIT and MIT-IBM Watson AI Lab · USA

Models that label suspicious chains of transfers by looking at who sent the money at the start of the chain and who received it at the end. A human makes the decision on a disputed operation; automatic blocking without review is not acceptable.

  • Classifying transfer chains as ordinary or needing review
  • Finding new suspicious chains in a large flow of operations
  • Preparing material for the compliance team
Sizes
checkpoints from 0.1 to 45 MB
Hardware
from: Laptop
Commercial use allowedDetails
FinanceGGUFNot maintained2023

AdaptLLM Finance

AdaptLLM · not disclosed

Finance-tuned versions of Llama 2: reading industry texts, reports and questions about terminology. Not investment advice: decisions are made by a specialist.

  • Reading financial news and reports
  • Answering questions about financial terminology
  • A base for fine-tuning to your own financial task
Sizes
7B и 13B
Hardware
from: Laptop
Commercial use with conditionsDetails
FinanceNot maintained2020

FinBERT

Prosus · Netherlands

A classic model that determines the tone of financial news: positive, negative or neutral. English only, runs fast on a CPU.

  • Scoring the tone of company news
  • Labeling reports and press releases
  • Signals for analytics dashboards
Sizes
110M
Hardware
from: Laptop
Commercial use allowedDetails

Voice assistants 6

Voice assistants2026

Samsone

Samsung · South Korea

Tiny audio-understanding models for smartphones: they listen to speech, music and ambient sounds and answer in text - describing a recording and answering questions about it. They run on the device itself; prompts and answers are in English - no other languages are present in the training data.

  • Describing an audio recording in words
  • Answering questions about a sound
  • Identifying the type of sound and the setting
Sizes
99M – 356M
Hardware
from: Laptop
Non-commercial onlyDetails
Voice assistants2024–2026

Audio Flamingo

NVIDIA · USA

Models that listen to speech, sounds and music and answer questions about them. Audio Flamingo Next handles recordings up to 30 minutes. Research use only.

  • Detailed descriptions of audio recordings
  • Questions and answers about a long recording
  • Tagging music and sounds
Sizes
0.5B – 8B
Hardware
from: Laptop
Non-commercial onlyDetails
Voice assistantsGGUF2025–2026

MiniCPM-o

OpenBMB (ModelBest, Tsinghua University) · China

A small model that sees, hears and replies by voice in real time, and can clone a voice. Voice dialogue in English and Chinese, text in 30+ languages.

  • Voice assistant on your own server
  • Analyzing videos and documents
  • Voice answers about a camera image
Sizes
8B – 9B
Hardware
from: Laptop
Commercial use allowedDetails
Voice assistants2024–2026

Moshi / Hibiki

Kyutai · France

A voice assistant that listens and speaks at the same time, with no delay for recognition and synthesis. Hibiki does simultaneous speech-to-speech translation between several European languages.

  • Real-time voice conversation partner
  • Simultaneous speech translation
  • Zero-latency voice interfaces
Sizes
2B – 7B
Hardware
from: Laptop
Commercial use with conditionsDetails
Voice assistantsRUGGUF2025

Qwen Omni

Alibaba (Qwen) · China

Models that understand text, images, audio and video and reply by voice in real time. Qwen3-Omni speaks 10 languages, including Russian.

  • Voice assistant for customers
  • Analyzing calls and videos
  • Voice answers about documents and images
Sizes
3B – 30B-A3B
Hardware
from: Laptop
Commercial use allowedDetails
Voice assistantsRUNot maintained2023

SeamlessM4T / Seamless

Meta · USA

Speech and text translation across roughly a hundred languages, including Russian: speech to text, speech to speech, and streaming translation that keeps intonation.

  • Speech-to-speech translation
  • Translating and transcribing recordings
  • Streaming translation
Sizes
281M – 2.3B
Hardware
from: Laptop
Non-commercial onlyDetails

Avatars 5

Avatars2025

LatentSync

ByteDance · China

Matches lip movements in an existing video to a new voice track. Version 1.6 works at 512 pixels and produces a sharper face.

  • Dubbing videos into another language with lip sync
  • Editing lines in finished video without reshooting
  • Talking avatars for training courses
Sizes
requires 8–18 GB of VRAM
Hardware
from: Laptop
Commercial use with conditionsDetails
Avatars2024–2025

MuseTalk

Tencent Music (Lyra Lab) · China

Real-time lip sync: matches the mouth in a video to new audio. Suits video translation and live avatars.

  • Dubbing video into another language
  • Live avatar in a video chat
  • Editing speech in a finished video
Sizes
under 1B
Hardware
from: Laptop
Commercial use allowedDetails
AvatarsNot maintained2024

LivePortrait

Kuaishou (Kling) · China

Animates a portrait from a reference video: an actor's facial expressions and head turns are transferred to the photo. Runs fast even on a weak GPU.

  • Animating portraits
  • Transferring an actor's expressions to a character
  • Mascot animation
Sizes
under 1B
Hardware
from: Laptop
Commercial use allowedDetails
AvatarsNot maintained2023

SadTalker

Xi'an Jiaotong University and Tencent AI Lab · China

An older lightweight talking-head model: one photo plus audio becomes a video. Runs on weak hardware, but quality is noticeably below newer models.

  • Talking photo for greetings
  • Simple voiced avatars
Sizes
under 1B
Hardware
from: Laptop
Commercial use allowedDetails
AvatarsNot maintained2020

Wav2Lip

IIIT Hyderabad · India

The classic lip-to-audio sync model, still popular in hobbyist setups. Lip movements are accurate but the face looks blurry; the license is non-commercial.

  • Quick dubbing tests
  • Comparison with newer lip-sync models
  • Educational and research projects
Sizes
small model, 96-pixel face
Hardware
from: Laptop
Non-commercial onlyDetails

Satellite and geo 5

Satellite and geo2025–2026

Ai2 OlmoEarth

Allen Institute for AI (Ai2) · USA

Ai2's family of models for Sentinel-1, Sentinel-2 and Landsat imagery, with ready-made fine-tunes for mangroves, deforestation and ecosystem types. The license excludes the extractive industries.

  • Monitoring deforestation and forest condition across the supply chain
  • Classifying land and crops from image series
  • Image embeddings for finding similar plots
Sizes
Nano – Large (Base about 114M)
Hardware
from: Laptop
Commercial use with conditionsDetails
Satellite and geo2025

IBM–ESA TerraMind

IBM and the European Space Agency (ESA) · USA / Europe

A multimodal Earth model: understands optical and radar imagery, terrain, vegetation index and land use maps, and can generate a missing data type (for example, a "see-through-clouds" image from radar).

  • Analyzing fields and forests even in cloudy weather using radar imagery
  • Land use maps for assessing plots
  • Flood and wildfire assessment (ready-made fine-tunes available)
Sizes
tiny – large (checkpoints from ~200 MB to ~3.8 GB)
Hardware
from: Laptop
Commercial use allowedDetails
Satellite and geo2023–2025

IBM–NASA Prithvi

IBM and NASA · USA

Foundation models for Landsat and Sentinel-2 satellite imagery that account for image time series. Ready-made fine-tunes for floods, burn scars and crop types, plus a separate weather model, WxC.

  • Mapping crops and field condition over the season
  • Assessing flood zones and burn scars after natural disasters
  • Monitoring changes in buildings and land use
Sizes
tiny – 600M (imagery), 2.3B (Prithvi WxC weather)
Hardware
from: Laptop
Commercial use allowedDetails
Satellite and geo2025

MBZUAI TerraFM

MBZUAI · UAE

A compact research model for satellite imagery, trained on both optical (Sentinel-2) and radar (Sentinel-1) data. Narrower in scope and community than Prithvi and TerraMind.

  • Classification and segmentation of satellite imagery after fine-tuning
  • A base for a land monitoring prototype
Sizes
TerraFM-B (ViT-Base)
Hardware
from: Laptop
Commercial use allowedDetails
Satellite and geoNot maintained2023–2024

Ai2 SatlasPretrain

Allen Institute for AI (Ai2) · USA

Pretrained models from the Satlas project for Sentinel-2, Landsat and high-resolution aerial imagery. The predecessor of OlmoEarth, still used in TorchGeo.

  • Detecting objects in imagery: solar farms, wind turbines, ships
  • Mapping roads and buildings from aerial photos
  • A starting point for fine-tuning your own geo model
Sizes
Swin-v2 and ResNet backbones (Base)
Hardware
from: Laptop
Commercial use allowedDetails

Image generation 4

Image generationGGUF2024–2025

SANA

NVIDIA · USA

NVIDIA's fast image model: 4K images in seconds, runs even on a laptop GPU. The Sprint version generates in 1–2 steps.

  • Bulk image generation
  • High-resolution visuals
  • Real-time generation inside apps
Sizes
0.6B – 4.8B
Hardware
from: Laptop
Commercial use allowedDetails
Image generationGGUF2022–2024

Stable Diffusion

Stability AI · UK

The model that started open image generation. A huge ecosystem of fine-tunes, styles and plugins; runs even on a home PC. The popular SDXL-Lightning and Hyper-SD accelerators were made by ByteDance.

  • Illustrations and banners for advertising
  • Backgrounds and scenes for product cards
  • Fine-tuning to a brand style
Sizes
0.9B – 8B
Hardware
from: Laptop
Commercial use with conditionsDetails
Image generationNot maintained2023–2024

ControlNet

Lvmin Zhang (Stanford) and the community · USA

An add-on for image models: sets pose, outlines, depth or floor plan so the result follows the required composition exactly.

  • Image from a sketch or outline
  • Keeping pose and composition
  • Interior visualization from a floor plan
Sizes
0.4B – 1.3B
Hardware
from: Laptop
Commercial use allowedDetails
Image generationNot maintained2023–2024

PixArt

Huawei Noah's Ark Lab and partners · China

A compact 0.6B image model with quality on par with much larger ones. The Sigma version does 4K; suits modest hardware.

  • Illustrations for articles and social media
  • Backgrounds for product cards
  • Quick visual drafts
Sizes
0.6B
Hardware
from: Laptop
Commercial use allowedDetails

3D 3

3D2024–2026

DUSt3R / MASt3R

NAVER LABS Europe · France (NAVER, South Korea)

The family that started "single-pass" 3D reconstruction from a pair or set of photos without camera calibration. MASt3R added point matching and scale; MUSt3R and BLASt3R added video support.

  • 3D scene from several photos without calibration
  • Point matching between images
  • Mapping from video (SLAM)
Sizes
0.57B – 0.69B
Hardware
from: Laptop
Non-commercial onlyDetails
3D2024–2026

TripoSR / TripoSG

VAST (TripoSR together with Stability AI) · China

VAST family: a 3D model from a single photo. TripoSR runs in under a second, TripoSG gives cleaner geometry, TripoSplat builds a scene from Gaussian points.

  • 3D product model from a photo
  • Object assets for games and AR
  • Quick 3D prototype for printing
Sizes
up to 1.5B
Hardware
from: Laptop
Commercial use allowedDetails
3DNot maintained2022–2023

Point-E / Shap-E

OpenAI · USA

Early open OpenAI models that create a 3D object from text or an image in seconds. Quality is basic, but they are fast and easy to run.

  • Rough 3D mock-ups from a description
  • Quick object prototypes for games and AR
  • Training and research pilots in 3D
Sizes
40M – 1B
Hardware
from: Laptop
Commercial use allowedDetails

Faces 3

Faces2021–2026

InsightFace

InsightFace (deepinsight) · China

The most widely used open toolkit for face detection and recognition. Many identity-preserving image generators are built on it. The pretrained weights are non-commercial.

  • Detecting and comparing faces in photos
  • Face-based access in prototypes
  • Face processing as part of other AI systems
Sizes
packages from 16 MB to 407 MB
Hardware
from: Laptop
Non-commercial onlyDetails
Faces2025

LVFace

ByteDance · China

Transformer-based face recognition from ByteDance, one of the most accurate open models on benchmarks. Weights are published in ONNX format but are non-commercial.

  • Comparing faces and searching a photo database
  • Research on recognition accuracy
  • Access control prototypes
Sizes
ViT-T – ViT-L
Hardware
from: Laptop
Non-commercial onlyDetails
FacesNot maintained2023–2024

IP-Adapter-FaceID

Tencent AI Lab (h94) · China

One of the first adapters that transfer a face from a photo into a generated image. The SD 1.5 versions run on low-end cards, but the weights are non-commercial.

  • Portraits from a photo in different styles
  • Image series with one character
  • Avatar experiments
Sizes
adapters for SD 1.5 and SDXL
Hardware
from: Laptop
Non-commercial onlyDetails

Visual document search 3

Visual document searchGGUF2026

EVIE

Tencent · China

Finds the right pages in documents as pictures, without turning them into text: matches a query against how the page looks. Version 8B scores 66.75 on ViDoRe V3. The index can be compressed to 32 vectors per page.

  • searching scans, slides and reports without text recognition
  • RAG over documents with tables and charts
  • searching archives of contracts and catalogs
Sizes
4.5B и 8B
Hardware
from: Laptop
Commercial use allowedDetails
Visual document search2025

ModernVBERT / ColModernVBERT

Illuin Technology, EPFL, CentraleSupélec · France

A compact (250M) model for searching document pages as images. According to the authors, it matches models 10 times larger and runs without a GPU.

  • Search across scans and PDFs on a modest server
  • Indexing document archives
  • Search across slides and manuals
Sizes
250M
Hardware
from: Laptop
Commercial use allowedDetails
Visual document search2024–2025

ColPali / ColQwen

Illuin Technology (ViDoRe team) · France

Searches PDFs and scans as images: pages do not need to be OCR'd first, the model finds the right one for a question directly, including tables and charts. Trained on English.

  • Search across scans, presentations and PDFs
  • RAG over documents with tables and charts
  • Search across technical documentation
Sizes
256M – 3B
Hardware
from: Laptop
Commercial use with conditionsDetails

Video 2

Video2022–2024

RIFE (Practical-RIFE)

hzwer (Zhewei Huang) and co-authors · China

Generates intermediate frames: turns 24–30 fps into 60 fps and more and makes smooth slow motion. Versions 4.24+ smooth out video from generative models well.

  • Increasing video frame rate
  • Smooth slow-motion video
  • Smoothing clips from AI generators
Sizes
lightweight model (size not stated on the model card)
Hardware
from: Laptop
Commercial use allowedDetails
VideoNot maintained2023–2024

AnimateDiff

Shanghai AI Lab and CUHK · China

A module that brings Stable Diffusion image models to life, turning them into short animations. One of the first open video technologies.

  • Short animations in brand style
  • Animated covers and banners
  • Animated stickers
Sizes
motion module on top of SD 1.5 / SDXL
Hardware
from: Laptop
Commercial use allowedDetails

Robotics 1

Robotics2025

SmolVLA

Hugging Face · USA

A small robot control model that runs on a regular laptop. Trained on open data from the LeRobot community, suited to low-cost robot arms.

  • Controlling a low-cost robot arm
  • Quick robotization pilots and demos
  • Training staff and students
Sizes
450M
Hardware
from: Laptop
Commercial use allowedDetails

Virtual try-on 1

Virtual try-on2024–2025

CatVTON / CatV2TON

Sun Yat-sen University and Pixocial · China

A lightweight try-on model that runs on a regular GPU. There is a mask-free version and CatV2TON, which also tries clothes on in video.

  • Trying a garment on a customer's photo
  • Draft product cards on a model
  • Try-on in a short video
Sizes
899M
Hardware
from: Laptop
Non-commercial onlyDetails

Weather and climate 1

Weather and climateNot maintained2023

Huawei Pangu-Weather

Huawei Cloud · China

One of the first weather neural networks, published in Nature and added to ECMWF charts. The weights are open for research only; commercial use is prohibited.

  • Research weather forecasts a week ahead
  • Comparison with other weather models on your own data
  • Training courses on weather neural networks
Sizes
4 models of ~1.1 GB each (1, 3, 6 and 24-hour steps)
Hardware
from: Laptop
Non-commercial onlyDetails

Biology and chemistry 1

Biology and chemistry2022–2026

ESM (ESM-2, ESM3, ESM C, ESMFold2)

EvolutionaryScale / Chan Zuckerberg Biohub (ESM-2 — Meta AI) · USA

Protein language models: they understand amino acid sequences, predict structure (ESMFold2) and help with protein design. Since 2026 all open versions are under MIT.

  • Protein embeddings for predicting properties (stability, solubility)
  • Predicting 3D structures of proteins and complexes
  • Screening enzyme and antibody design candidates before lab work
Sizes
8M – 15B (ESM-2), 300M – 6B (ESM C), 1.4B (open ESM3)
Hardware
from: Laptop
Commercial use allowedDetails

Autonomous driving 1

Autonomous driving2020–2026

comma.ai openpilot (supercombo)

comma.ai · USA

An open driver assistance system: a neural network keeps the lane and controls speed from a camera, plus a driver attention monitoring model. The models live right in the repository and are updated constantly.

  • A research testbed for driver assistance systems
  • Studying driver attention monitoring with an in-cabin camera
  • Comparison with your own lane-keeping algorithms
Sizes
compact, designed for an in-vehicle device
Hardware
from: Laptop
Commercial use allowedDetails

Where people usually hit the wall

The bottleneck here is speed. A language model on a CPU answers in seconds rather than instantly, so live customer chat is usually not the place for it. On overnight jobs the difference is invisible. One more point: you need roughly as much RAM as the model file weighs, plus headroom for context.

By hardware

Need a model for your task?

An open model can run on your own server: data stays in-house, there is no per-request fee, and the model can be fine-tuned on your documents.

  1. SelectThe model and size for your task and hardware budget
  2. DeployOn your server or in a closed network, with an API
  3. Fine-tuneOn your data, or connect a knowledge base
  4. IntegrateInto your CRM, ERP, bot, website or team chat
Discuss deployment