Hugging Face Open Models: Model List, Licenses and Articles

About Hugging Face

Hugging Face runs the platform for sharing models and datasets itself, but it also builds and releases its own models. This page covers two of its organizations: HuggingFaceTB (Hugging Face Smol Models Research), which studies small models, and HuggingFaceM4, its multimodal team working on vision-language models.

It stands out for releasing small models together with how they were trained and on what data. The organization card also points to its guide to training small models (The Smol Training Playbook) and high-quality pre-training datasets.

What It Releases

  • SmolLM3 (July 2025): according to the model card, a 3B language model that pushes the limits of small models, with dual-mode reasoning (thinking on or off), 6 languages and long context, and “fully open", down to its training details.
  • SmolVLM and SmolVLM2 (2025): small 256M, 500M and 2.2B models that also read images and video.
  • Idefics (HuggingFaceM4, 2023 to 2024): vision-language models from 8B to 80B.
  • There are also intermediate training checkpoints and 100M-class experimental models.

Licenses

SmolLM, SmolVLM and Idefics2 and 3 are Apache-2.0. The early Idefics (9B, 80B) use a custom license (other). Read each model’s license text before use. This summary is not legal advice.

Running Them Locally

  • Most models are small enough to try even on a CPU-only laptop. SmolLM3 has an official ONNX build, and the earlier SmolLM2 has official GGUF builds.
  • For recent models, use third-party GGUF conversions.

Sources: the organization cards of HuggingFaceTB and HuggingFaceM4; the model card of HuggingFaceTB/SmolLM3-3B; the model lists of both organizations on Hugging Face (all as of 2026-09-26).

Our Coverage and Data

Local Model Watch has published 0 article(s) on Hugging Face’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 HuggingFaceTB, HuggingFaceM4
Models on Hugging Face 118
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 118 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.

HuggingFaceTB (84 models)

→ Scroll horizontally to see all columns

Released Model Task Parameters License Official variants
2026-04-24 HuggingFaceTB/nanowhale-100m text generation 110M apache-2.0 base
2026-04-03 HuggingFaceTB/SmolLM3-3B-GSM8K-SFT text generation 3.1B apache-2.0 —
2026-03-25 HuggingFaceTB/qwen3-1.7b-gsm8k-sft text generation 1.7B apache-2.0 —
2025-07-21 HuggingFaceTB/SmolLM3-3B-checkpoints — — apache-2.0 —
2025-07-08 HuggingFaceTB/SmolLM3-3B text generation 3.1B apache-2.0 ONNX, Base
2025-04-14 HuggingFaceTB/SmolVLM2-2.2B-Base vision-language 2.2B apache-2.0 —
2025-02-27 HuggingFaceTB/SmolLM2-135M-intermediate-checkpoints — — apache-2.0 —
2025-02-27 HuggingFaceTB/SmolLM2-360M-intermediate-checkpoints — — apache-2.0 —
2025-02-26 HuggingFaceTB/SmolLM2-1.7B-intermediate-checkpoints — — apache-2.0 —
2025-02-11 HuggingFaceTB/SmolVLM2-500M-Video-Instruct vision-language 507M apache-2.0 —
2025-02-11 HuggingFaceTB/SmolVLM2-256M-Video-Instruct vision-language 256M apache-2.0 —
2025-02-09 HuggingFaceTB/SmolVLM2-2.2B-Instruct vision-language 2.2B apache-2.0 —
2025-01-20 HuggingFaceTB/SmolVLM-500M-Instruct vision-language 507M apache-2.0 —
2025-01-18 HuggingFaceTB/SmolVLM-256M-Instruct vision-language 256M apache-2.0 —
2025-01-13 HuggingFaceTB/SmolVLM-500M-Base vision-language 507M — —

HuggingFaceM4 (34 models)

→ Scroll horizontally to see all columns

Released Model Task Parameters License Official variants
2024-08-06 HuggingFaceM4/Idefics3-8B-Llama3 vision-language 8.5B apache-2.0 —
2024-06-21 HuggingFaceM4/Florence-2-DocVQA vision-language 823M — —
2024-05-03 HuggingFaceM4/Meta-Llama-3-8B-tokenizer-pad-is-eos — — — —
2024-05-02 HuggingFaceM4/idefics2-8b-chatty vision-language 8.4B apache-2.0 AWQ
2024-04-23 HuggingFaceM4/tiny-random-Llama3ForCausalLM text generation 4M — —
2024-04-10 HuggingFaceM4/idefics2-8b vision-language 8.4B apache-2.0 base-AWQ, AWQ, base
2024-03-29 HuggingFaceM4/tiny-random-vmistral-m4-split-lm_head text generation 2M — —
2024-03-10 HuggingFaceM4/idefics2_raven_finetuned text generation 8.4B apache-2.0 —
2024-01-09 HuggingFaceM4/VLM_WebSight_finetuned text generation 8.2B apache-2.0 —
2023-07-12 HuggingFaceM4/idefics-9b text generation 8.9B other instruct
2023-07-06 HuggingFaceM4/idefics-80b text generation 80.0B other instruct
2023-01-19 HuggingFaceM4/tiny-random-vopt-clip text generation — bigscience-openrail-m —

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.