Text2026
Ricursive Intelligence · USA
A dense (not MoE) model with 65B parameters, pretrained on 12 trillion tokens, with a context of up to 1 million tokens. It is built for programming and agents: by the authors' own test it scores 80.52 on Terminal-Bench 2.1 and beats much larger American open models (Reflection Beam 501B with 80.1, Meta Muse Spark Open 816B with 76.4, NVIDIA Nemotron 3 Ultra 550B with 56.4). The authors started from the Olmo 3 recipe and claim a 16 to 55 times gain in training efficiency on code. It was designed for NVIDIA graphics cards. BF16 and NVFP4 are published for the instruction-tuned version, plus a BF16 base. This is a preview release, and the figures are the authors' own, not independent.
- an agent for programming in the terminal and a repository
- long tasks with a context of hundreds of thousands of tokens
- your own tuning on top of an open base model
- Sizes
- 65B (dense)
- Hardware
- from: Cluster
TextGGUF2025–2026
Moonshot AI · China
Very large Moonshot MoE models for agentic work. K3 (2.8 trillion parameters) was the largest open model at release, with up to 1M tokens of context and image understanding; K2.7-Code is built for programming. On November 4 the weights of K3.1 Pro were opened (2.8 trillion parameters, 104B active, 1M tokens of context, images): a refinement of K3 for programming, 3D and financial research; by the tests of Moonshot itself, Terminal-Bench 4.0 is 65.7%. The weights are in MXFP4, about 1.56 TB, run through vLLM and SGLang, and need a serious cluster.
- Multi-step agents: search, data collection, reports
- In-depth document analysis
- Help for developers
- Sizes
- 16B-A3B – 2.8T-A104B
- Hardware
- from: 1 GPU
TextOllama2024–2026
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
Medicine2026
Hugging Face (HuggingFaceBio) · USA
An open 14.8B language model for working with DNA, built on Qwen3-14B. You send a raw sequence right in an ordinary chat, and it answers in English: it judges whether a mutation is benign or pathogenic, finds regions that control gene activity, assigns a sequence to an organism group, predicts the next bases and explains its reasoning. DNA and text go in one stream, with DNA encoded in groups of six bases. By the authors' tests, on five sets it scores 66.7% without reasoning, 27 points above Qwen3-14B, and costs about 1,800 times less than GPT-6.1 Sol at a close result. It fits on a single 80 GB card, so sensitive data can stay with you. Viral data was excluded from training because of the risks. The conclusions are computed and do not replace laboratory verification.
- assessing DNA variants: benign or pathogenic
- finding regions that control gene activity
- a research assistant for biologists on your own server
- Sizes
- 14.8B
- Hardware
- from: 1 GPU
3D2026
NVIDIA · USA
A diffusion model that creates three-dimensional body animation from a text description and set conditions: poses, positions of hands and feet, a route and points to pass through. There are versions for the SOMA skeleton (30 joints), the Unitree G1 robot (34 joints) and the SMPL-X human (22 joints). Version 1.2 is trained on a corrected set: in 1.1 some test recordings leaked into training, so the new version is the one to compare against other models. According to the authors, foot skating is reduced and text following is slightly weaker. The data of some variants is proprietary (Bones Rigplay), while the open variants are trained on Bones-SEED.
- animating characters for games and video from a description
- motions for humanoid robots (the Unitree G1 skeleton)
- motion drafts with a given route and poses
- Sizes
- not stated
- Hardware
- from: 1 GPU
Robotics2025–2026
NVIDIA · USA
"World" models for robots and self-driving vehicles: they generate realistic video of physical scenes and predict actions. Cosmos 3 combines understanding, generation and control. In October the surgical versions came out: Cosmos3-H-Embodied-Nano (16B, one model for predicting video, actions and policy on surgical robots) and Cosmos3-H-Dreams (a real-time world for simulating operations with the Versius robot). They are under the NVIDIA Open Model License.
- Synthetic video for training robots and self-driving vehicles
- Testing scenarios in simulation
- Robot control (Policy versions)
- Sizes
- 2B – 65B
- Hardware
- from: 1 GPU
Weather and climate2024–2026
Allen Institute for AI (Ai2) · USA
A fast climate model emulator: simulates the atmosphere years and decades ahead on a single GPU. Coupled with an ocean model (SamudrACE) for long-term scenarios. The new HiRO-ACEv2 adds a diffusion model that sharpens the coarse 100 km grid to 3 km: precipitation, wind and pressure every 6 hours for decades ahead, worldwide or for a chosen region. By the estimate of Ai2, global 3 km data come out about 300 times cheaper than with the X-SHiELD model. For now it reproduces only the X-SHiELD climate, not real weather from the archive.
- Decades-long climate scenarios to assess long-term asset risks
- Large-scale what-if runs on temperature and precipitation
- Preparing data for crop yield and energy demand models
- Sizes
- checkpoint of about 1.8 GB
- Hardware
- from: 1 GPU
Text2026
Reflection AI · USA
An open model from the American company Reflection with 501B parameters (23B active, a 1 million token context), text only, mainly English. Stated as economical in reasoning: according to the authors it needs about three times fewer tokens for thinking at comparable quality. Strong in programming, tool calling and science. Weights in BF16, FP8 and NVFP4. The benchmark figures are the authors' own.
- an agent for solving tasks in a repository and a terminal
- long tasks with tools and search (via MCP)
- solving scientific and mathematical problems with reasoning
- Sizes
- 501B-A23B
- Hardware
- from: Cluster
Text2026
Meta Superintelligence Labs · USA
Meta's largest open model: 816B parameters, 42B active per token, a 1 million token context. It understands text, images, video and audio and answers in text. Built for programming and long agentic tasks. Trained on data in more than 100 languages. Access to the weights on Hugging Face is granted manually.
- an agent for working with code and long tasks with tool calling
- analyzing screenshots, charts, documents, video and recordings
- generating training data for other models
- Sizes
- 816B-A42B
- Hardware
- from: Cluster
Image generation2026
Sperid Labs · not stated
Generates images directly in pixels, without a compressed latent space. After fine-tuning, the same weights estimate depth from a single photo and restore and upscale images 4 times. Prompts are in English, the default size is 1024 by 1024.
- generating images from a description
- estimating scene depth from a single photo
- restoring and upscaling old or compressed photos
- Sizes
- 3B
- Hardware
- from: 1 GPU
Text2026
Shanghai Artificial Intelligence Laboratory · China
An agent model for long tasks: searching and gathering evidence, research, development, preparing reports and presentations, analyzing vulnerabilities in an authorized environment. A preview based on GLM-5.2, 256K-token context, text only. An FP8 version exists.
- deep search and research on a topic with a verifiable result
- developing programs and preparing reports and presentations
- checking and fixing vulnerabilities in an authorized environment
- Sizes
- 744B (MoE, based on GLM-5.2)
- Hardware
- from: Cluster
Medicine2026
NVIDIA · USA
A language model for computed tomography: looks at a whole chest or abdominal CT volume, not separate slices, writes structured reports and answers questions with reasoning. English only.
- a draft CT report for a physician
- yes-or-no answers about a scan with reasoning
- research in medical imaging
- Sizes
- 4B (based on Qwen3.5-4B with a 3D encoder)
- Hardware
- from: 1 GPU
Medicine2026
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
Computer-use agents2026
Beijing Academy of Artificial Intelligence (BAAI) · China
An agent model of 27B (based on Qwen3.8-27B) for long tasks: it proposes a solution, measures the result, analyzes errors and revises the solution over several rounds. A 262K-token context. Stated as a research model.
- iterative improvement of code and algorithms
- machine learning tasks with a verifiable result
- deep search and gathering information on a topic
- Sizes
- 27B
- Hardware
- from: 1 GPU
Text2026
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
Tabular data2026
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
Text to speechRU2024–2026
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
Text analysisRU2026
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
Autonomous driving2020–2026
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
3D2024–2026
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
MedicineGGUF2023–2026
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
TextOllama2023–2026
Shanghai AI Laboratory · China
Models from Shanghai AI Laboratory. The early InternLM line is general-purpose; the new Intern-S1/S2 is scientific: it understands formulas, molecules, charts and images.
- Research assistant: papers, formulas, data
- Analysis of scientific and technical documents
- Corporate chat on small models
- Sizes
- 1.8B – about 1T
- Hardware
- from: Laptop
Documents and OCR2026
Jina AI · Germany
Document parsing in a single model: a whole page becomes Markdown - text in correct reading order, tables and formulas in LaTeX. Built on DeepSeek-OCR, with only 0.6B of its 3.4B parameters active.
- Converting scans and PDFs to Markdown
- Recognizing tables and formulas
- Parsing invoices, acts and reports
- Sizes
- 3.4B-A0.6B
- Hardware
- from: 1 GPU
Text2024–2026
MBZUAI, Institute of Foundation Models (IFM, LLM360 project) · UAE
Fully open models from the UAE: data, training code and intermediate checkpoints are published along with the weights. K2-Horizon (2026) spans 0.9B to 375B with context up to 512K tokens.
- Reasoning, maths and technical questions
- Analysing long documents
- Agents and writing code
- Sizes
- 0.9B – 375B-A23B
- Hardware
- from: Laptop
Image + text2026
Apple · USA
Reads long documents differently: it turns pages into downscaled images and expands the full text only for the pages needed to answer. A long document then costs far less context.
- Answering questions over multi-page documents
- Finding the right page in a large report or case file
- Working with scans and PDFs without a separate OCR step
- Sizes
- 9B
- Hardware
- from: 1 GPU
Math and reasoning2026
Paradigma · not disclosed
A 1B model that does exactly one thing: it solves maths problems. It answers in a single turn, with no dialogue, and handles non-maths questions poorly. A companion Value Model scores someone else's solution step by step.
- Solving maths problems in a single answer
- Picking the best solution out of several candidates
- Scoring reasoning step by step when training another model
- Sizes
- 1B
- Hardware
- from: 1 GPU
TextGGUF2025–2026
Renmin University of China (GSAI) and Ant Group (inclusionAI) · China
Diffusion language models: text is written in blocks and then refined rather than word by word, which speeds up generation. LLaDA2.2 can edit what it has written and targets agents. LLaDA-Image is a separate product.
- Fast generation of code and text
- Agent scenarios with long context
- Research into alternatives to standard LLMs
- Sizes
- 8B – 100B (MoE)
- Hardware
- from: 1 GPU
Music and soundGGUF2022–2026
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
TextGGUF2025–2026
Xiaomi · China
Xiaomi models for reasoning and agents: from the compact MiMo-7B to MiMo-V2.6-Pro with 1.02 trillion parameters. The larger versions understand text, images, video and audio, with a 1M token context. Languages: English and Chinese.
- Logic and calculation tasks
- Agents with tools
- Help for developers
- Sizes
- 7B – 1,02T-A42B
- Hardware
- from: Laptop
Voice assistants2026
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
TextRU2026
SberDevices (ai-forever) · Russia
A Russian and English research prototype: the model writes text in blocks at once (diffusion) rather than word by word, which speeds up responses. The authors do not recommend it for production systems.
- Experiments with faster generation
- Fine-tuning small models for your own tasks
- Research
- Sizes
- 0.6B – 4B
- Hardware
- from: Laptop
Finance2024–2026
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
Drawings and CAD2026
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
Robotics2025–2026
GigaAI · China
A robot control model trained mostly on synthetic data from a world model. It reduces spending on collecting data from real robots.
- Controlling a robot arm
- Fine-tuning on a small amount of your own data
- Sorting and assembly pilots
- Sizes
- 3.5B
- Hardware
- from: 1 GPU
Weather and climate2023–2026
Google DeepMind · UK
Google DeepMind's family of global weather models: GraphCast (10-day forecast), GenCast (probabilistic ensemble) and WeatherNext 2 with cyclone forecasting. Since August 2026 the weights are cleared for commercial use.
- Medium-range weather forecasts for planning shifts, voyages and deliveries
- Probabilistic assessment of extreme weather for insurance portfolios
- Tropical cyclone track forecasts for marine and port operations
- Sizes
- from lightweight 1° versions to full 0.25°
- Hardware
- from: 1 GPU
Acoustic monitoring2021–2026
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
Computer vision2024–2026
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
Autonomous driving2025–2026
NVIDIA · USA
Vision-language-action models for self-driving vehicles: they plan a trajectory from camera video and explain the decision in text. Used to develop and test autopilot systems, not as a ready-made autopilot.
- Auto-labeling camera recordings to train your own driver assistance systems
- Analyzing complex road scenes with text explanations
- Testing autopilot systems in simulation on rare scenarios
- Sizes
- 10B – 34B
- Hardware
- from: 1 GPU
Biology and chemistry2022–2026
AlQuraishi Lab (Columbia University) and the OpenFold consortium · USA
A fully open reproduction of AlphaFold 2 and then AlphaFold 3 under Apache 2.0, with training data. OpenFold3 predicts complexes of proteins, nucleic acids and ligands.
- Predicting structures of proteins and ligand complexes
- Fine-tuning on the company's own data (training code is open)
- An in-house structural analysis service without sending data outside
- Sizes
- a single set of weights per version
- Hardware
- from: 1 GPU
Autonomous driving2026
Alibaba (Qwen team) · China
An autonomous driving model based on Qwen3.5-4B: 3D detection of objects around the vehicle, answers to questions about the road scene and trajectory planning in one model.
- A perception and planning prototype for autonomous vehicles on closed sites
- Answering questions about camera recordings when reviewing incidents
- Labeling road scenes to train your own models
- Sizes
- 4B
- Hardware
- from: 1 GPU
Deepfake detection2023–2026
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
TextRUGGUF2025–2026
Swiss AI (ETH Zurich, EPFL, CSCS) · Switzerland
Switzerland's public open model: weights, data and recipe are open, with more than 1000 languages in training. Version 1.5 understands images.
- Multilingual assistant
- Answers based on documents
- Analysis of images and scans (v1.5)
- Sizes
- 0.5B – 70B
- Hardware
- from: Laptop
Acoustic monitoring2023–2026
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
TextOllama2024–2026
LG AI Research · South Korea
Korean-English models from LG. Most of the line is non-commercial, but the flagship K-EXAONE 2.0 with 750 billion parameters is released under Apache 2.0.
- Corporate assistant
- Working with Korean and English texts
- Analysis of documents and images (4.5)
- Sizes
- 1.2B – 750B-A37B
- Hardware
- from: Laptop
Tabular data2026
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
Robotics2026
Ant Group (Robbyant) · China
A robot control model from Ant Group trained on a large volume of data from real robots. Version 2.0 works with different types of robot arms.
- Controlling a two-armed robot
- Fine-tuning for your own operation
- Assembly and sorting pilots
- Sizes
- 4B – 6B
- Hardware
- from: 1 GPU
Weather and climate2024–2026
Microsoft Research · USA
A foundation model of Earth's atmosphere: global weather forecasts, plus separate versions for air quality and ocean waves. Computes a forecast in seconds instead of hours on a supercomputer.
- Your own forecast of temperature, wind and precipitation for company locations
- Sea state estimates for planning voyages and port operations
- Air pollution forecasts for industrial sites
- Sizes
- about 1.3B (a small test version is available)
- Hardware
- from: 1 GPU
Weather and climate2022–2026
NVIDIA · USA
NVIDIA's set of weather and climate models: global FourCastNet forecasts, downscaling to kilometers (CorrDiff), regional storm forecasts (StormCast), climate generation (cBottle, Atlas). Run via Earth2Studio.
- Global forecasts followed by downscaling to the region you need
- Short-term forecasts of thunderstorms and heavy rain for dispatch services
- Generating many weather scenarios for stress tests
- Sizes
- 98M – 2.5B (FourCastNet 3 — about 711M)
- Hardware
- from: 1 GPU
Image + textGGUF2024–2026
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
Robotics2026
Xiaomi · China
Open robot control models from Xiaomi. Robotics-1 is designed for household and kitchen tasks, U0 combines scene understanding and action.
- Controlling a robot arm by command
- Household and service scenarios
- Base for fine-tuning
- Sizes
- 4B – 5B
- Hardware
- from: 1 GPU
Satellite and geo2025–2026
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
Image generationGGUF2026
Ideogram · Canada
Open weights of the Ideogram model, known for precise typography. Under a non-commercial license: for business, suitable only for testing.
- Testing text-in-image generation
- Research and prototypes
- Sizes
- about 9B
- Hardware
- from: 1 GPU
Image + text2023–2026
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
Math and reasoningOllama2025–2026
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
Acoustic monitoring2019–2026
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
Tabular data2026
Google Research · USA
Google's large tabular model: classification and regression from examples without training, with numeric and categorical columns. Weights are for non-commercial use only.
- Pilot comparison with current scoring models
- Exploring customer data
- Testing churn hypotheses
- Sizes
- about 1.6B
- Hardware
- from: 1 GPU
Biology and chemistry2022–2026
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
Image + textOllama2023–2026
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
MedicineOllama2023–2026
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
Documents and OCRGGUF2025–2026
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
Robotics2025–2026
Ai2 (Allen Institute for AI) · USA
A fully open robot control model that first "reasons" about space and trajectory, then acts. Its reasoning can be checked.
- Controlling a robot arm with explainable steps
- Fine-tuning for your own robot
- Research pilots
- Sizes
- 5B – 8B
- Hardware
- from: 1 GPU
3D2025–2026
Meta and the University of Oxford (VGG) · USA / UK
Reconstructs a 3D scene from one, several or hundreds of photos in seconds: camera positions, depth and a point cloud. Best Paper at CVPR 2025.
- 3D model of a room or object from a photo series
- Camera pose estimation for photogrammetry
- Point cloud for measurements and comparison with the plan
- Sizes
- about 1.2B
- Hardware
- from: 1 GPU
Drawings and CADGGUF2026
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
Voice assistants2024–2026
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
Fact-checking and judgesRU2025–2026
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
CodeOllama2025–2026
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
Acoustic monitoring2024–2026
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
Math and reasoningGGUF2025–2026
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
MedicineGGUF2025–2026
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
Image + textRUGGUF2023–2026
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
Video2025–2026
Skywork AI (Kunlun) · China
An interactive "world model": generates video of a game world in real time and responds to keyboard and mouse input. Version 3.0 keeps scene memory for minutes.
- Game world prototypes without an engine
- Interactive demos and simulations
- Generating data to train agents
- Sizes
- 1.8B – 17B
- Hardware
- from: 1 GPU
Image + text2024–2026
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
Robotics2025–2026
NVIDIA · USA
NVIDIA's foundation model for humanoid robots and robot arms: it sees, understands a command and outputs movements. Built into the Isaac ecosystem.
- Controlling humanoid robots
- Fine-tuning for your own robotic cell
- Training in simulation with transfer to a real robot
- Sizes
- 2B – 3B
- Hardware
- from: 1 GPU
Text to speechRU2026
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
Tabular data2025–2026
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
Text analysisRU2026
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
Biology and chemistry2024–2026
Arc Institute (with Together AI, Stanford, NVIDIA) · USA
DNA language models with context up to a million nucleotides: they assess the impact of mutations, annotate genomes and generate sequences. Evo 2 is trained on genomes from all domains of life.
- Assessing the likely harmfulness of genetic variants for research
- Annotating genomes of microorganisms and plants
- Finding promising sequences in breeding and synthetic biology
- Sizes
- 1B – 40B
- Hardware
- from: 1 GPU
Math and reasoningGGUF2026
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
Code2026
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
MedicineGGUF2025–2026
Ant Healthcare (Ant Group) and Zhejiang Provincial Medical Information Center · China
A large medical MoE model based on Ling-flash-2.0: 100B parameters with 6B active, so it answers quickly. Does not replace a doctor; decisions are made by a specialist.
- Reference answers to staff on clinical questions
- Draft discharge summaries for a doctor to review
- Searching medical literature
- Sizes
- 100B-A6B
- Hardware
- from: 1 GPU
Medicine2023–2026
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
Weather and climate2024–2026
European Centre for Medium-Range Weather Forecasts (ECMWF) · Europe (intergovernmental organization)
ECMWF's weather neural network running operationally: a 15-day forecast four times a day, an ensemble version with 51 scenarios, and since version 2, ocean waves.
- Running your own forecast from open initial data
- Ensemble forecasts to estimate the probability of frost, downpours and storms
- Wave forecasts for marine operations
- Sizes
- checkpoint of about 1 GB
- Hardware
- from: 1 GPU
TextGGUF2024–2026
Prime Intellect · USA
Models trained in a distributed way on GPUs from around the world. INTELLECT-3 (106B) is further trained with reinforcement learning for math, code and agents.
- Reasoning and math tasks
- Programming help
- Agents with tool calling
- Sizes
- 10B – 106B-A12B
- Hardware
- from: 1 GPU
MedicineGGUF2025–2026
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
MedicineOllama2025–2026
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
TextOllama2024–2026
Allen Institute for AI (Ai2) · USA
Fully open models: not only the weights but also the data, training code and intermediate checkpoints are published. Useful when transparent provenance matters.
- Assistant and answers based on documents
- Reasoning tasks (Think versions)
- Fine-tuning on your data with a clear model history
- Sizes
- 1B – 32B
- Hardware
- from: Laptop
Image + text2023–2025
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
Computer visionGGUF2024–2025
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
3D2025
Meta and Carnegie Mellon University · USA
A single model builds a metric 3D reconstruction from photos, and uses camera, depth or pose data when available. One weights variant is under Apache 2.0.
- 3D reconstruction of an object or room from photos
- Exporting the scene to COLMAP format for further processing
- Depth and camera pose estimation
- Sizes
- about 1.2B
- Hardware
- from: 1 GPU
Speech to textRU2025
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
3D2025
Shanghai AI Lab · China
Reconstructs a 3D scene and camera positions from a set of photos or a video without relying on a "reference" frame. Pi3X gives smoother point clouds and approximate scale in meters.
- 3D scene reconstruction from video
- Camera pose estimation from frames
- Point clouds for research and prototypes
- Sizes
- 0.96B – 1.4B
- Hardware
- from: 1 GPU
Speech to textRU2024–2025
Mihaj Dolgushin · not disclosed
About two dozen Whisper, wav2vec2 and WavLM fine-tunes for Karelian, including recordings where the speaker switches from Karelian to Russian. The model cards are empty, so quality must be checked on your own recordings.
- Transcribing field and archive recordings in Karelian
- Subtitles for Karelian-language material
- Handling recordings that mix Karelian and Russian
- Sizes
- from 300M to 1B across the different models
- Hardware
- from: Laptop
Math and reasoningGGUF2024–2025
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
Satellite and geo2025
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
Computer vision2023–2025
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
Acoustic monitoring2025
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
3DGGUF2025
Meta · USA
Reconstructs the 3D shape of an object or a human body from one ordinary photo, even when the object is partly hidden. Two models: Objects and Body.
- 3D model of an item from a catalog photo
- Estimating body pose and shape from a photo
- Try-on and AR scenarios
- Sizes
- size not stated on the model card
- Hardware
- from: 1 GPU
Deepfake detection2025
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
Biology and chemistry2024–2025
MIT (Jameel Clinic) and Recursion · USA
An open MIT-licensed alternative to AlphaFold 3: predicts structures of protein, DNA and small-molecule complexes; Boltz-2 estimates binding strength, BoltzGen designs new binding proteins.
- Predicting how a candidate molecule binds to a target protein
- Ranking compounds by predicted binding strength before synthesis
- Designing binder proteins for a given target
- Sizes
- checkpoints of about 2 GB
- Hardware
- from: 1 GPU
Recommenders2025
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
Documents and OCRGGUF2025
Ai2 (Allen Institute for AI) · USA
A model and toolkit for converting PDFs into clean text at scale, preserving reading order, tables and formulas. Built to process millions of pages.
- Bulk digitisation of a PDF archive
- Converting contracts and reports into text
- Preparing documents for search and RAG
- Sizes
- 7B
- Hardware
- from: 1 GPU
Satellite and geo2023–2025
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
Drawings and CAD2025
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
Robotics2025
Physical Intelligence · USA
Robot control models from Physical Intelligence: folding laundry, tidying up, handling objects. π0.5 copes better in unfamiliar settings.
- Controlling robot arms and two-armed robots
- Fine-tuning for your own operations
- Pilots for automating manual work
- Sizes
- about 3B
- Hardware
- from: 1 GPU
Computer vision2023–2025
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
Text2024–2025
xAI · USA
xAI publishes the weights of previous Grok generations. The models are very large and need a GPU cluster, so in practice they are rarely run.
- Research on large models
- Assistant on your own infrastructure
- Text generation and analysis
- Sizes
- 314B (Grok-1), Grok-2 is larger
- Hardware
- from: Cluster
Math and reasoningGGUF2025
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
Search and RAGRU2023–2025
lingtrain · not disclosed
A set of LaBSE fine-tunes that turn sentences into vectors: Buryat, Mari, Udmurt, Ingush, Chuvash, Sakha and Kalmyk. Trained on pairs with Russian; quality differs by language.
- Aligning parallel texts in Russian and a national language
- Semantic search across bilingual archives
- Building training sets for translation
- Sizes
- LaBSE fine-tunes; parameter count is not stated on the cards
- Hardware
- from: Laptop
Faces2025
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
Text analysisRUGGUF2025
Johns Hopkins University (JHU CLSP) · USA
A multilingual encoder trained on more than 1800 languages. The list includes Tatar, Bashkir, Chuvash, Udmurt, Buryat, Komi, Ingush and other languages of Russia. A base for classifiers and search over such texts.
- Classifying requests and documents in national languages
- Semantic search over texts in rare languages
- Extracting names, dates and titles from text
- Sizes
- 140M and 307M
- Hardware
- from: Laptop
Acoustic monitoring2025
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
Fact-checking and judges2024–2025
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
Math and reasoningOllama2025
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
MedicineGGUF2025
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
Medicine2025
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
Math and reasoning2025
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
Deepfake detection2024–2025
Xiaohongshu, USTC and Shanghai Jiao Tong University · China
An AI-image detector made of several experts: some look at visual artifacts, others at noise. The hard Chameleon benchmark was released with it. It errs in both directions - a human reviews the result.
- Checking realistic AI images without obvious artifacts
- Comparing detectors on hard examples
- Fine-tuning for your own type of content
- Sizes
- several experts based on ConvNeXt and CLIP
- Hardware
- from: 1 GPU
Drawings and CAD2025
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
Image + textGGUF2025
Moonshot AI · China
An efficient MoE vision model (16B, 3B active) with a long context and a reasoning version. Handles long documents and video well.
- Analysing long PDFs and presentations
- Answering questions about video
- Operating interfaces from screenshots
- Sizes
- 16B-A3B
- Hardware
- from: 1 GPU
Satellite and geo2025
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
Robotics2025
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
Drawings and CAD2024–2025
DFKI and RPTU Kaiserslautern · Germany
It builds a sequence of CAD modelling commands straight from a text description of a part, from a beginner phrase to a designer-level wording. The output is a draft; an engineer checks and signs off the document.
- A draft part from a text description
- Quickly trying out shape options at an early stage
- Filling a library of standard parts
- Sizes
- size not stated on the model card
- Hardware
- from: 1 GPU
Image generationGGUF2025
ByteDance Seed · China
A unified model that understands images, generates them and edits them in a conversation. Similar to how images work in ChatGPT.
- Photo editing in a conversation
- Answering questions about an image
- Image generation with explanations
- Sizes
- 14B-A7B
- Hardware
- from: 1 GPU
TextOllama2025
DeepSeek · China
A reasoning model that thinks step by step before answering. Strong at calculations, logic and code; compact distilled versions are available.
- Complex calculations and logic checks
- Analysis of contracts and internal policies
- Help for developers
- Sizes
- 1,5B – 671B
- Hardware
- from: Laptop
Acoustic monitoring2024–2025
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
Finance2025
The Fin AI · international project
Models that spell out their reasoning before answering financial questions with numbers and tables. Not investment advice: decisions are made by a specialist.
- Calculation questions on statements with the working shown
- Working through tasks with tables and numbers from documents
- Checking calculations made by hand
- Sizes
- 8B и 14B
- Hardware
- from: 1 GPU
Fact-checking and judgesRU2024–2025
NVIDIA · USA
Large NVIDIA scorers for selecting and fine-tuning answers. The multilingual GenRM version lists Russian among its languages. The scorer itself makes mistakes and does not replace manual review on important tasks.
- Choosing the best of several candidate answers
- Preparing data to fine-tune your own model
- Scoring assistant answers in Russian and other languages
- Sizes
- 49B, 70B and 340B
- Hardware
- from: Cluster
Acoustic monitoring2024–2025
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
Math and reasoningGGUF2024–2025
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
Acoustic monitoring2024–2025
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
Drawings and CAD2024–2025
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
Text to speechGGUF2024–2025
Shanghai Jiao Tong University and partners · China
A voice cloning model that needs only a few seconds of a sample, in English and Chinese. The community has released many fine-tuned versions for other languages, including Russian.
- Voice cloning
- Voicing audiobooks and videos
- Research and prototypes
- Sizes
- about 340M
- Hardware
- from: Laptop
Math and reasoningGGUF2025
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
Visual document search2025
Nomic AI · USA
Search across PDF pages and scans as images. The cards list English, Italian, French, German and Spanish — Russian is not among them.
- Search across an archive of scans and PDFs
- Search across tables and diagrams inside documents
- Picking pages for an AI assistant answer
- Sizes
- 3B and 7B
- Hardware
- from: 1 GPU
Video2024–2025
HPC-AI Tech · Singapore
A fully open video generation project: weights, code and training recipe. Version 2.0 at 11B makes video from text and from an image.
- Video from a text description
- Animating images
- Training your own video model
- Sizes
- up to 11B
- Hardware
- from: 1 GPU
Math and reasoningOllama2024–2025
Qwen (Alibaba) · China
Qwen's first open reasoning model: it thinks step by step before answering and comes close to DeepSeek-R1 on maths tasks with only 32B parameters.
- Calculations and formula checks
- Complex analytics with step-by-step breakdowns
- Checking the logic of contracts and internal policies
- Sizes
- 32B
- Hardware
- from: 1 GPU
Math and reasoningGGUF2025
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
Video2025
StepFun · China
A large 30B video model producing clips of up to 204 frames. Needs server hardware, but is open under MIT.
- Video from a description
- Animating images
- Sizes
- 30B
- Hardware
- from: Cluster
Medicine2024–2025
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
Computer-use agentsGGUF2025
Microsoft Research · USA
An agent model that plans actions both in an interface (buttons on screen) and for a robot (arm movements). For now more of a research base than a finished product.
- Pilots in interface control
- Research projects spanning screens and robotics
- Analyzing screenshots with an action plan
- Sizes
- 8B
- Hardware
- from: 1 GPU
Robotics2024–2025
Stanford, Berkeley and partners · USA
The first large open vision-language-action model: a robot arm carries out commands like "put the apple in the bowl". OFT makes it several times faster.
- Controlling a robot arm by text command
- Pilots for robotizing simple operations
- Base for fine-tuning to your own robot
- Sizes
- 7B
- Hardware
- from: 1 GPU
Math and reasoningGGUF2025
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
TextOllama2023–2025
Ai2 · USA
Ai2 fine-tunes of Llama with a fully open recipe: data, code and all intermediate stages. Tulu 3 405B is one of the largest openly fine-tuned models; OLMo chat versions use the same recipe.
- Employee assistant on your own server
- Math and precise instruction following
- Reference recipe for your own fine-tuning
- Sizes
- 7B – 405B
- Hardware
- from: Laptop
Fact-checking and judges2025
Atla · UK
An 8B judge model: it scores another model answer against your criteria and writes a rationale. The judge itself makes mistakes and does not replace manual review on important tasks.
- Scoring chatbot answers against your own criteria
- Comparing two versions of a prompt or model
- Filtering out weak answers before they reach a person
- Sizes
- 8B
- Hardware
- from: 1 GPU
Medicine2024–2025
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
Math and reasoningOllama2024–2025
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
Medicine2024
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
Music and sound2024
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
Drawings and CAD2023–2024
Manycore Research · China
It assembles a 3D model of a product from three standard drawing projections - front, top and side - including cases where the lines in the drawing are incomplete. The output is a draft; an engineer checks and signs off the document.
- Turning a three-view drawing into a 3D model
- Digitising an archive of paper drawings for furniture and cabinets
- Checking that the views agree with each other
- Sizes
- size not stated on the model card, several checkpoints published for different inputs
- Hardware
- from: 1 GPU
Biology and chemistry2024
Google DeepMind and Isomorphic Labs · UK
The reference model for the structure of biomolecules and their complexes. Weights are provided for non-commercial research only; companies need commercial access via Google Cloud or open alternatives (Boltz, OpenFold3).
- Academic research on protein and complex structures
- Benchmarking open alternatives against the reference on your own targets
- Sizes
- a single set of weights
- Hardware
- from: 1 GPU
TextOllama2024
Nexusflow · USA
Fine-tuned Llama 3 and Qwen 2.5 models from Nexusflow. Athene-V2-Agent is specially trained for function calling and agent scenarios. Commercial use is prohibited.
- Research on agents and function calling
- Comparison with commercial models
- Experiments with a chat assistant
- Sizes
- 70B – 72B
- Hardware
- from: 1 GPU
Video2024
Zhipu AI (Z.ai) and Tsinghua University · China
A 2–5B video model that runs on a single gaming GPU. A popular base for research and add-ons.
- Short clips from text
- Animating images
- Video fine-tuning experiments
- Sizes
- 2B – 5B
- Hardware
- from: 1 GPU
Finance2023–2024
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
Music and sound2023–2024
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
Image + textGGUF2024
Google · USA
Google's vision model built on Gemma, designed as a base for fine-tuning on a narrow task: captions, object detection, reading text.
- Fine-tuning for your own recognition task
- Finding objects in photos
- Reading text in images
- Sizes
- 3B – 28B
- Hardware
- from: Laptop
Finance2023–2024
AI4Finance Foundation · USA
An open set of lightweight add-ons for ordinary language models that work with financial texts and news. Not investment advice: decisions are made by a specialist.
- Assessing the tone of financial news and reports
- Tagging mentions of companies and instruments in text
- Preparing digests from a stream of business news
- Sizes
- adapters for 6B - 20B base models
- Hardware
- from: 1 GPU
Visual document search2024
TIGER-Lab · Canada
Turns an image-plus-text model into an embedding model: one vector for a page, a diagram or a captioned photo. The card states English.
- Search across a mixed archive of texts and images
- Search across document pages as images
- Finding similar cards and illustrations
- Sizes
- about 4B (based on Phi-3.5-V)
- Hardware
- from: 1 GPU
Visual document search2024
University of Waterloo, Tevatron project · Canada
Searches page screenshots: the page is not OCRed but turned into a single vector, so the index is more compact than with late-interaction models. The card lists English and French.
- Search across scans and PDFs without OCR
- Search across presentations and reports with complex layouts
- Picking pages for an AI assistant answer
- Sizes
- 2B (based on Qwen2-VL)
- Hardware
- from: 1 GPU
TranslationRUNot maintained2023–2024
Johns Hopkins University and Microsoft · USA
Research translators based on Llama 2. The first ALMA covered 5 pairs with English, including Russian; X-ALMA expanded coverage to 50 languages.
- Translation between English and Russian
- Experiments with LLM-based translation
- Base for fine-tuning a translator
- Sizes
- 7B – 13B
- Hardware
- from: 1 GPU
Image + textNot maintained2023–2024
Hugging Face · France / USA
Open vision models from Hugging Face that reproduced the closed Flamingo. Idefics3 became the basis for the compact SmolVLM line.
- Answering questions about images
- Analysing documents and screenshots
- A base for fine-tuning
- Sizes
- 8B – 80B
- Hardware
- from: 1 GPU
MedicineNot maintained2024
Paige · USA
Paige's pathology foundation model, trained on millions of digital slides. The first version is Apache 2.0, the second is for research only. Does not replace a doctor; decisions are made by a specialist.
- Slide patch features for research classifiers
- Selecting slides for re-review in research projects
- Comparison with other pathology models on your own archive
- Sizes
- about 632M
- Hardware
- from: Laptop
MedicineGGUFNot maintained2023–2024
M42 Health · UAE
Clinical models from Abu Dhabi-based M42, built on Llama and tuned to answer medical questions. Does not replace a doctor; decisions are made by a specialist.
- Reference answers to staff on clinical questions
- Draft discharge summaries and letters for a doctor to review
- Searching medical literature
- Sizes
- 8B – 70B
- Hardware
- from: Laptop
Tabular dataNot maintained2024
ML Foundations · USA
A foundation model for predictions on tables: it classifies and forecasts from a handful of examples, with no separate task-specific training.
- Classifying table rows from a few examples
- Predicting a value from a data row
- Quickly testing hypotheses on new datasets
- Sizes
- 8B
- Hardware
- from: 1 GPU
Text analysisRUNot maintained2024
udmurtNLP · not disclosed
A project around the Udmurt language: encoders based on mBERT, rubert-tiny2 and Glot500, part-of-speech taggers, a typo correction model and Udmurt-English embeddings.
- Parsing Udmurt texts by part of speech
- Correcting typos and recognition errors
- Semantic search over Udmurt material
- Sizes
- 33M – 187M
- Hardware
- from: Laptop
Fact-checking and judgesNot maintained2024
RLHFlow · USA
An answer scorer that returns a breakdown across several attributes rather than a single overall score. The scorer itself makes mistakes and does not replace manual review on important tasks.
- Choosing the best of several candidate answers
- Preparing data for model fine-tuning
- Scoring assistant answers across several attributes
- Sizes
- 8B
- Hardware
- from: 1 GPU
RecommendersNot maintained2024
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
Deepfake detectionNot maintained2024
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
TextOllamaNot maintained2023–2024
01.AI · China
Bilingual (English and Chinese) 01.AI models of 6–34B, with versions supporting up to 200K tokens of context. No new open releases since 2024.
- Chat assistant on a single GPU
- Analysis of long documents
- Classification and data extraction from text
- Sizes
- 6B – 34B
- Hardware
- from: Laptop
MedicineGGUFNot maintained2024
Avignon University and Nantes University · France
Mistral 7B fine-tuned on PubMed Central papers, plus several merges with the general model. Compact and easy to run. Does not replace a doctor; decisions are made by a specialist.
- Searching and summarising medical papers
- Draft reference materials for staff
- Explaining medical terminology
- Sizes
- 7B
- Hardware
- from: Laptop
Text analysisRUNot maintained2023–2024
CIS, LMU Munich · Germany
Detects which language a text is written in: the third version covers more than two thousand labels, including Tatar, Bashkir, Chuvash, Udmurt, Mari, Erzya, Komi and other languages of Russia.
- Sorting mixed text archives by language
- Filtering out noise and foreign languages before training
- Routing requests to the right operator or model
- Sizes
- a FastText model; parameter count is not stated on the model card
- Hardware
- from: Laptop
Virtual try-onNot maintained2024
KAIST and OMNIOUS.AI · South Korea
One of the best-known open virtual try-on models: moves a garment from a product photo onto a photo of a person, keeping prints and logos well. Non-commercial license.
- Pilot of a fitting room on a store website
- Prototype product cards on a model without a photo shoot
- Comparison with commercial try-on services
- Sizes
- based on SDXL
- Hardware
- from: 1 GPU
Text analysisNot maintained2024
University of Southern Denmark · Denmark
Turns a free-form occupation description into a standard HISCO code in 13 languages. Built for historical archives, but also useful for cleaning up job title reference lists. A human makes the decision about a candidate; automatic screening without review must not be used.
- Mapping mixed occupation names onto a single code
- Processing archives of HR and statistical data
- Preparing data for reporting
- Sizes
- based on CANINE-s, size not stated on the model card
- Hardware
- from: Laptop
MedicineGGUFNot maintained2024
Saama AI Research · India
Llama 3 fine-tuned on medical and biological data. One of the first strong open medical models of 2024. Does not replace a doctor; decisions are made by a specialist.
- Extracting data from medical documents
- Draft discharge summaries for a doctor to review
- Searching medical literature
- Sizes
- 8B, 70B
- Hardware
- from: Laptop
RecommendersNot maintained2024
VinAI Research · Vietnam
A 7B language model further trained on review and product-card text for recommendation and rating-prediction tasks. The licence forbids commercial use; a store still needs its own order history.
- Predicting the rating a user would give an item
- Suggesting the next purchase from history
- Researching text-based recommendations
- Sizes
- 7B
- Hardware
- from: 1 GPU
TextRUGGUFNot maintained2024
Tweeties (Francois Remy, Ghent University, and co-authors) · Belgium
A language model for Tatar: Mistral 7B converted to a Tatar tokenizer, plus a version for translating between Tatar and a dozen other languages, including Russian. Check quality on your own texts.
- Generating and continuing Tatar text
- Translating between Tatar and Russian
- A base for fine-tuning to your own task
- Sizes
- 7B
- Hardware
- from: Laptop
TextOllamaNot maintained2023–2024
Hugging Face (H4) · USA
Hugging Face educational chat models based on Mistral, Gemma and Mixtral with an open fine-tuning recipe. Zephyr 7B Beta showed a small model can be trained to large-model level without human labeling.
- Lightweight chat assistant
- Reference and starting point for your own fine-tuning
- Drafts of texts and replies
- Sizes
- 7B – 141B-A35B
- Hardware
- from: Laptop
Deepfake detectionNot maintained2023–2024
Sichuan University and co-authors · China
An open model for finding forgeries in images: it outputs a pixel-level mask of altered regions. It errs in both directions; its map is a hint for an expert, not proof of a forgery.
- Finding pasted and erased fragments in photos
- Checking document scans for edits
- A baseline when comparing manipulation-localization models
- Sizes
- a Vision Transformer based model
- Hardware
- from: 1 GPU
Virtual try-onNot maintained2024
Xiao-i Research · China
An early popular open try-on model: one version for upper-body garments, another for full-length outfits. Non-commercial license.
- Trying tops on a model photo
- Full-length try-on: tops, bottoms, dresses
- Fitting room prototype for testing
- Sizes
- based on Stable Diffusion
- Hardware
- from: 1 GPU
TextOllamaNot maintained2023–2024
TinyLlama (SUTD researchers) · Singapore
A 1.1B model with the Llama 2 architecture, trained on 3 trillion tokens. Now behind newer small models, but still a popular base for experiments and fine-tuning.
- Simple chatbots on low-end hardware
- Experiments and team training
- A base for fine-tuning on a narrow task
- Sizes
- 1.1B
- Hardware
- from: Laptop
ForecastingNot maintained2024
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
Virtual try-onNot maintained2024
KAIST · South Korea
A research try-on model from CVPR 2024, one of the first built on Stable Diffusion. Now mostly used as a comparison baseline.
- Pilot of upper-body garment try-on
- Comparing quality of different try-on models
- Training your own try-on on the open code
- Sizes
- based on Stable Diffusion
- Hardware
- from: 1 GPU
Photo editingNot maintained2024
XPixel Group (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shanghai AI Lab and others) · China
Powerful SDXL-based restoration of badly damaged photos: it recreates details rather than just upscaling. Hardware-hungry; non-commercial license.
- Restoring old and blurry photos
- Upscaling with detail reconstruction
- Archive restoration pilots
- Sizes
- based on SDXL, plus LLaVA 13B for captions
- Hardware
- from: 1 GPU
Satellite and geoNot maintained2023–2024
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
RerankersGGUFNot maintained2023–2024
University of Waterloo, Castorini group · Canada
Rerankers that are language models: they receive the whole list of retrieved passages and reorder it as a list, instead of scoring passages one by one. Heavier than ordinary rerankers.
- Reordering a long list of search results
- Selecting sources for an AI assistant answer
- Research comparisons of retrieval approaches
- Sizes
- 7B – 13B
- Hardware
- from: 1 GPU
TextOllamaNot maintained2023–2024
WizardLM (Microsoft and Peking University) · USA / China
Fine-tunes of Llama, Mistral and StarCoder using Evol-Instruct, which automatically makes instructions more complex. WizardLM-2 was released in April 2024 and removed almost immediately, so only the 2023 versions are relevant.
- Complex multi-step instructions
- Help for developers
- Solving math problems
- Sizes
- 7B – 70B
- Hardware
- from: Laptop
RecommendersNot maintained2023
Renmin University of China · China
An approach where every item gets a short identifier code that a language model understands, and the model predicts that code. Delta weights for three categories are published; no licence is stated on the card.
- Suggesting the next item from history
- Turning a catalogue into short codes for the model
- Recommendations from a shopper text request
- Sizes
- about 7B (a delta to LLaMA weights)
- Hardware
- from: 1 GPU
TextOllamaNot maintained2023
Microsoft Research · USA
Microsoft research models based on Llama 2, trained to choose a reasoning approach for each task. The orca-mini model in Ollama is a different project by independent developer Pankaj Mathur.
- Research on reasoning methods
- Comparison with modern small models
- Training specialists
- Sizes
- 7B – 13B
- Hardware
- from: Laptop
Voice assistantsRUNot maintained2023
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
Tabular dataNot maintained2023
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
Text to speechRUNot maintained2023
Coqui · Germany
A popular model for cloning a voice from a short sample in 17 languages, including Russian. Coqui has shut down and development has stopped.
- Voice cloning from a sample
- Multilingual voiceover
- Research and prototypes
- Sizes
- about 470M
- Hardware
- from: Laptop
Documents and OCRNot maintained2023
Meta · USA
An early model that converts scientific PDFs into text with formulas. Now outdated and outperformed by almost all modern OCR models.
- Converting scientific papers from PDF into text with formulas
- Digitising technical documentation
- Sizes
- 250M – 350M
- Hardware
- from: Laptop
MedicineNot maintained2023
Bo Wang's lab (University of Toronto, Vector Institute) · Canada
An early medical model on Llama 2 70B, trained on dialogues based on medical texts. Now mainly of research interest. Does not replace a doctor; decisions are made by a specialist.
- Research pilots on medical dialogue
- Training materials for staff
- Comparison with newer medical models
- Sizes
- 70B
- Hardware
- from: 1 GPU
TextRUGGUFNot maintained2022–2023
Sber (ai-forever) · Russia
Sber's multilingual model covering 61 languages, including languages of the peoples of Russia and the CIS. Separate fine-tunes exist for Buryat, Yakut, Tatar, Bashkir, Kazakh and others, rare for open models.
- Texts in languages of the peoples of Russia and the CIS
- Base for fine-tuning on a less common language
- Drafts and templates in several languages
- Sizes
- 1.3B – 13B
- Hardware
- from: Laptop
TextRUGGUFNot maintained2023
SberDevices (ai-forever) · Russia
Separate mGPT versions, one per language: Bashkir, Buryat, Kalmyk, Mari, Ossetian, Tatar, Tuvan, Chuvash and Sakha. Each card also lists Russian and English.
- Generating and continuing text in a national language
- A base for fine-tuning to your own task
- Experiments with rare languages without training from scratch
- Sizes
- 1.3B
- Hardware
- from: Laptop
MedicineNot maintained2023
Shanghai Jiao Tong University and Shanghai AI Lab · China
An early general-purpose radiology model: understands 2D and 3D images (CT, MRI) together with text. More of a research base. Does not replace a doctor; decisions are made by a specialist.
- Research pilots on CT and MRI analysis
- Hints for doctors when reviewing images
- A base for fine-tuning on the clinic's own images
- Sizes
- size not stated on the model card
- Hardware
- from: 1 GPU
CybersecurityNot maintained2022–2023
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
TextOllamaNot maintained2023
LMSYS (Berkeley and partners) · USA
One of the first open chat models (2023): LLaMA fine-tuned on user conversations with ChatGPT. A historical milestone; today it is weaker than any modern model of the same size.
- Experiments and team training
- Simple chat assistant for tests
- Comparison with newer models
- Sizes
- 7B – 33B
- Hardware
- from: Laptop
Weather and climateNot maintained2023
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
Drawings and CADNot maintained2023
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
Photo editingNot maintained2023
S-Lab, Nanyang Technological University · Singapore
One of the first Stable Diffusion-based photo upscalers: restores realistic details. Non-commercial license.
- Upscaling photos with detail reconstruction
- Research restoration pilots
- Comparison with classic upscalers
- Sizes
- based on SD 2.1
- Hardware
- from: 1 GPU
Image generationNot maintained2023
DeepFloyd (Stability AI) · UK
An early model that was among the first to render text on images accurately. Today it is mostly of historical interest; development has stopped.
- Research experiments
- Prototype images with captions
- Sizes
- 0.4B – 4.3B
- Hardware
- from: 1 GPU
TextNot maintained2022–2023
EleutherAI · USA
Fully open models from the non-profit lab EleutherAI: GPT-NeoX-20B and the Pythia series with published intermediate training checkpoints.
- Base model for fine-tuning
- Research into model behavior
- Simple text generation and completion
- Sizes
- 70M – 20B
- Hardware
- from: Laptop
Acoustic monitoringNot maintained2022
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
TextNot maintained2022
Meta · USA
An early open Meta series matching GPT-3 in size. Outdated; useful for research and comparison.
- Research experiments
- Training specialists
- Comparison with modern models
- Sizes
- 125M – 66B (175B on request)
- Hardware
- from: Laptop
RecommendersNot maintained2022
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
TextGGUFNot maintained2022
BigScience (Hugging Face and the community) · France
One of the first large open models, trained by a community of hundreds of researchers in 46 languages. Today it is interesting mostly as a historical milestone.
- Text generation and translation in many languages
- Experiments and team training
- Base model for fine-tuning on a narrow task
- Sizes
- 560M – 176B
- Hardware
- from: Laptop
