Microsoft Open Models: Model List, Licenses and Articles
About Microsoft
Microsoft publishes 544 models on its microsoft organization on Hugging Face. Rather than large flagship production models, most are small and mid-size models with clear purposes from its research labs (Microsoft Research).
What It Releases
- Phi: a family of small language models. Phi-4-reasoning-vision-15B (January 2026) is a reasoning model that also reads images.
- Fara1.5 (July 2026): browser-operating agents (computer use agents). According to the model card, it comes from Microsoft Research AI Frontiers, looks at screenshots of the browser and outputs actions such as click, type and scroll. The 4B version is fine-tuned from Qwen3.5-4B; there is also a 27B version.
- Speech: VibeVoice-ASR-Streaming (September 2026), streaming speech recognition in 10 languages that identifies who said what, with custom hotwords.
- Image and video understanding: Mage-VL (July 2026), a 4B model that works natively on compressed video.
- BitNet: language models with 1.58-bit weights (such as bitnet-b1.58-2B-4T).
- The table also includes models such as the rho series, whose model cards do not describe their use.
Licenses
Most recent models are MIT (Phi, Fara1.5, VibeVoice, BitNet and others). Mage-VL is Apache-2.0, and research models based on Llama use the original Llama license. Some models require agreeing before access. Read each model’s license text before use. This summary is not legal advice.
Running Them Locally
- Official GGUF builds are limited to models such as phi-4, Phi-3-mini and bitnet-b1.58-2B-4T (as of 2026-09-26); for newer models, look for third-party conversions.
- With many small models, it is an easy publisher to try on consumer hardware, such as the task-focused Fara1.5-4B.
Sources: model cards of microsoft/Fara1.5-4B, microsoft/VibeVoice-ASR-Streaming-7B and microsoft/Mage-VL; the model list of the microsoft organization on Hugging Face (all as of 2026-09-26).
Our Coverage and Data
Local Model Watch has published 0 article(s) on Microsoft’s own models, 0 on third-party fine-tunes and quantized builds based on them, and 1 from its official blog. The model list below is collected by code from Hugging Face every day. Part of our publisher index; model families link to their family pages.
At a Glance
| Item | Value |
|---|---|
| Hugging Face organization | microsoft |
| Models on Hugging Face | 544 |
| Articles on its own models | 0 |
| Articles on derived models | 0 |
| Articles from its official blog | 1 |
Recent Models on Hugging Face
The publisher lists 544 models in total on Hugging Face. Below are up to 15 of the most recent generative models (text, image, video, audio and so on), the scope of this site, per organization. Classifiers, feature extractors, research components and repositories without a model card are omitted, and the publisher’s own quantized builds and helper variants are folded into the row of the original model. Parameter counts are left blank for repositories that pack weights into integer types, where Hugging Face’s count is not the real parameter count.
→ Scroll horizontally to see all columns
| Released | Model | Task | Parameters | License | Official variants |
|---|---|---|---|---|---|
| 2026-09-22 | microsoft/rho-roboeval | — | 5.2B | mit |
— |
| 2026-09-03 | microsoft/VibeVoice-ASR-Streaming-1.5B | audio | 2.8B | mit |
— |
| 2026-09-03 | microsoft/VibeVoice-ASR-Streaming-7B | audio | 8.7B | mit |
— |
| 2026-08-29 | microsoft/rho-libero | — | 5.2B | mit |
— |
| 2026-08-27 | microsoft/rho-yam-box | — | 5.2B | mit |
— |
| 2026-08-27 | microsoft/rho-ur-ai-trainer | — | 5.2B | mit |
— |
| 2026-08-27 | microsoft/rho-fr3-duo | — | 5.2B | mit |
— |
| 2026-07-25 | microsoft/Mage-VL | vision-language | 4.7B | apache-2.0 |
— |
| 2026-07-24 | microsoft/VibeVoice-ASR-BitNet | audio | 323M | mit |
— |
| 2026-07-18 | microsoft/Fara1.5-27B | vision-language | 27.4B | mit |
— |
| 2026-07-18 | microsoft/Fara1.5-4B | vision-language | 4.5B | mit |
— |
| 2026-07-15 | microsoft/bitnet-embedding-0.6b | — | — | mit |
— |
| 2026-07-15 | microsoft/bitnet-embedding-270m | — | — | mit |
— |
| 2026-07-02 | microsoft/HARC-Qwen2.5-7B-Instruct | text generation | 7.6B | apache-2.0 |
— |
| 2026-07-02 | microsoft/HARC-Llama-3.1-8B-Instruct | text generation | 8.0B | llama3.1 |
— |
From the Hugging Face API. Last changed 2026-09-27 (JST). Collected by code.
From Its Official Blog
Last updated 2026-09-27 (JST). The explanation at the top of this page was written with the help of AI from the primary sources it cites. The tables and lists under “Our Coverage and Data" are assembled by code from our article log and the Hugging Face API.