MiniMax Open Models: Model List, Licenses and Articles

September 26, 2026September 27, 2026

About MiniMax

MiniMax is an AI company that releases text, video and music generation models. Its MiniMaxAI organization on Hugging Face says it is “building artificial general intelligence with everyone, ensuring AI is accessible to everyone", and publishes technical reports alongside its models. It has only 21 models, but each is large and a flagship in its area.

What It Releases

• MiniMax-M3 (understanding text, images and video) is an MoE model with about 428B total and 23B active parameters. According to its model card, it is natively multimodal, trained on text, images and video together from the first step, with a one-million-token context. Its own sparse attention (MiniMax Sparse Attention) gives 9x faster prefill and 15x faster decoding than the previous M2 at a one-million-token context.
• MiniMax-H3 generates video with audio from images and text (our family page has a detailed explanation).
• MiniMax Music 3 generates complete songs up to five minutes long with vocals, from lyrics and a description of the music. According to its model card, it combines an 8B language model for the song’s overall structure with a 0.6B model for fine acoustic detail.
• Earlier families include the MiniMax-M2, M2.5 and M2.7 text models (228.7B total) and MiniMax-M1.

Licenses

Each model has its own license, and the terms differ a lot. Always read the license text.

• MiniMax-M3 (MiniMax Community License): free to use, with conditions on commercial use. You must prominently display “Built with MiniMax M3"; products and services with more than US$20 million in yearly revenue need prior written authorization from MiniMax, and below that a one-time notice to MiniMax is enough.
• MiniMax-H3 (MiniMax H3 Community License Agreement): the European Union, the United Kingdom, the Republic of Korea and the United States are excluded from the territory where the license applies. Use under this license is not permitted in those regions.
• MiniMax Music 3 (MiniMax-Music3 Community License): conditional on complying with applicable laws and the Acceptable Use Policy.
• Among earlier models, M2, M2.1 and M2.5 use a “Modified MIT" license, M2.7 has its own license, and M1 is Apache-2.0.

This summary is not legal advice.

Running Them Locally

• There are no official GGUF builds (as of 2026-09-26). Quantized builds come from third parties.
• M3 activates about 23B, but all of its roughly 428B parameters must be loaded. That is too large for consumer GPUs. M3 has an official MXFP8 build.
• The video and music models are split into files per component. The MiniMax Music 3 repository has separate language model, condition encoder and VAE files. For memory requirements, see the tables in our articles and family pages.

Sources: the MiniMaxAI organization card; model cards and license texts of MiniMaxAI/MiniMax-M3, MiniMaxAI/MiniMax-H3 and MiniMaxAI/MiniMax-Music3 (all as of 2026-09-26).

Our Coverage and Data

Local Model Watch has published 1 article(s) on MiniMax’s own models, 1 on third-party fine-tunes and quantized builds based on them, and 0 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

Recent Models on Hugging Face

The publisher lists 21 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.

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

Licenses of Its Own Models (as recorded in our articles)

The license can differ from model to model even within one publisher. This is not legal advice; check each model’s license text before use.

By Model Family

Articles on Its Own Models

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.