Meta Open Models: Model List and Licenses

September 26, 2026

About Meta

Meta is the publisher of Llama, the open-weight language models. On Hugging Face it releases Llama, Llama Guard and Prompt Guard through the meta-llama organization and research models through the facebook organization. This page covers both.

Llama is one of the models that got local LLMs going, and it is still the base of many fine-tunes and quantized builds.

What It Releases

• The newest family on meta-llama is Llama 4, from April 2025. It has two models, Llama 4 Scout (109B total, 17B active, 16 experts) and Llama 4 Maverick (400B total, 17B active, 128 experts), both MoE models trained on images from the start. According to the model card, the context length is 10 million tokens for Scout and 1 million for Maverick.
• No newer Llama family has appeared on the meta-llama organization since then (as of 2026-09-26). That is why the meta-llama models in the table below stop at April 2025.
• Research models continue on the facebook organization, such as MobileMoE from June 2026 (MoE text models from 1.3B to 5.3B), but many use non-commercial licenses (fair-noncommercial-research-license, cc-by-nc-4.0).

Licenses

• Llama 4 uses its own Llama 4 Community License. Commercial use is allowed, but if your products or services had more than 700 million monthly active users on the Llama 4 release date, you must request a separate license from Meta. When distributing products or models built with Llama, you must prominently display “Built with Llama" and follow the Acceptable Use Policy.
• Access is gated on Hugging Face. meta-llama models can only be downloaded after you accept the license and the Acceptable Use Policy and request access.
• This summary is not legal advice; read the license text before use.

Running Them Locally

• Llama 4 activates 17B, but all the weights must be loaded. Even Scout is 109B in total, so it needs heavy quantization to fit on a single consumer GPU. If you want a small model, earlier generations such as the Llama 3.2 1B and 3B are options.
• GGUF builds of Llama come from third parties. There are no GGUF files on the meta-llama organization; for llama.cpp, Ollama and LM Studio, use conversions by quantizers.

Sources: the meta-llama organization card; the model card of meta-llama/Llama-4-Scout-17B-16E-Instruct and the Llama 4 Community License; the model lists of the meta-llama and facebook organizations on Hugging Face (all as of 2026-09-26).

Our Coverage and Data

Local Model Watch has not published an article on Meta’s models yet. 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

Recent Models on Hugging Face

The publisher lists 2429 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.

meta-llama (70 models)

facebook (2359 models)

From the Hugging Face API. Last changed 2026-09-26 (JST). Collected by code.

Last updated 2026-09-26 (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.