TextMath and reasoning

K2 (K2-Think, K2-V2, K2-Horizon)

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.

Developer
MBZUAI, Institute of Foundation Models (IFM, LLM360 project), UAE
First release
May 2024
Latest release
Sep 2026
Sizes
0.9B – 375B-A23B
License
Commercial use allowedApache 2.0
Russian
Not supported
Running
On your own serverAlso runs without a GPU
Industries
Science and research, Education, Software development
Deploy on your server Hugging Face GitHub

What it does

  • Reasoning, maths and technical questions
  • Analysing long documents
  • Agents and writing code
  • Verifiable training of your own models on open data

Where it is used

Research groupsIT departmentsEducation

Hardware requirements

LaptopLaptop or regular PC, up to 8 GB of VRAM — smaller versions
fits
1 GPUOne GPU with 16–80 GB — mid-size versions
fits
ClusterServer with several GPUs — flagship versions
fits

Versions

  1. K2-Horizon (0.9B, 3.7B, 7B, 32B, MoVA 36B-A4B, 375B-A23B)
  2. K2-Think-V2
  3. K2-V2 70B
  4. K2-Think 32B
  5. K2-65B

How to run it

On your own serverWeights are downloaded from Hugging Face and served with vLLM (load and API) or llama.cpp (modest hardware). The model then runs in a closed network with no per-request fees.Hugging Face
How much hardware you needCalculate the VRAM for the model size, context length and number of concurrent requests.Open the hardware calculator

I can set this up end to end: pick the model size, deploy it on your server and connect it to your systems.

Frequently asked questions

Can K2 (K2-Think, K2-V2, K2-Horizon) be used in a commercial project?

Yes. License: Apache 2.0. It allows commercial use, but it is still worth having a lawyer review the license before launch.

What hardware does K2 (K2-Think, K2-V2, K2-Horizon) need?

At minimum: Laptop or regular PC, up to 8 GB of VRAM — smaller versions. Some versions also run on an ordinary CPU, without a GPU. You can calculate the exact VRAM for your model size and context in the hardware calculator.

Does K2 (K2-Think, K2-V2, K2-Horizon) support Russian?

No. The model card lists its languages and Russian is not among them.

Where can I download K2 (K2-Think, K2-V2, K2-Horizon) and what does it cost?

The K2 (K2-Think, K2-V2, K2-Horizon) weights are open and free to download. You only pay for the hardware it runs on and for the setup. Source links are at the bottom of this page.

How I deploy it for clients

  1. SelectionI pick the model size for your task and hardware and test it on your examples.
  2. DeploymentI deploy it on your server or in a closed network and provide an API.
  3. Fine-tuningI fine-tune it on your data (LoRA) or connect a knowledge base — whichever is cheaper for the task.
  4. IntegrationI connect it to your CRM, ERP, bot, website or team chat and set up monitoring.

Similar models

TextOLMoAllen Institute for AI (Ai2) · USACommercial use allowed

Fully open models: not only the weights but also the data, training code and intermediate checkpoints are published. Useful when transparent provenance matters.

DetailsTextJaisInception (G42), MBZUAI and Cerebras · UAECommercial use allowed

A model family for Arabic and English, including Gulf dialects. Suits companies working with Arabic-speaking customers and government bodies in the region.

DetailsTextFalconTechnology Innovation Institute (TII) · UAECommercial use with conditions

A family from Abu Dhabi: from the early Falcon 40B and 180B to hybrid Falcon-H1 and tiny Falcon-H1-Tiny models of 90–600M parameters for devices.

Details

Source: huggingface.co/IFM/K2-Horizon-375B-A23B. Data checked against the model card on 22 Sep 2026. Have a lawyer review the license before commercial launch.