Official model card · read November 9, no generations by this site
MiniMax · Released 2026-08-02
MiniMax H3
An open-weights joint audio-video model: the official card puts the H3-Omni-Transformer at 33B dense single-stream (~13B AdaLN not loaded at inference), generating video with native 32 kHz stereo audio up to 15 seconds, 24 fps, up to 2K, plus an H3-Context-IR input-understanding module. The community license applies outside the US, EU, UK and South Korea. AA / Pi are text boards and do not apply; Physics-IQ scores video continuation (v2v), a different track from text- or image-to-video, and its rows read on 2026-11-07 contain no H3 — with no same-caliber score it stays off the boards and unrated.
Original model explanation ↗No comparable leaderboard score is included yet. Explore the evaluations and examples below.
Evaluations: conditions and results
Published primary-source account, not our run
The community compression HyperFlow H3 Flash, in three changes →
Singapore video company Video Rebirth compressed MiniMax's open-weights joint audio-video model H3 into a few-step version: drop 3 of 50 blocks, low-rank-factor 102 linear layers, quantize 180 layers to rotated W8A8 INT8, and add two-time conditioning for 4–8-step CFG-free sampling.
License text · US/EU/UK/KR need separate authorization
The community license: territory and Model Derivative ↗
How this capability connects to works and uses
Third-party skill-repo measurements · pinned at v0.11.0
The H3 prompt dialect: turns, speech onset and voice binding ↗
Public third-party repo · source material for this site's application case
The third-party skill repo drama-skills: the working repo for H3 prompting ↗
MiniMax model evolution
- MiniMax M3.1 Flash
The formal model reaches API, M Plan and MiniMax Code with 1M context and five effort settings; weights and Pro follow separately.
Original release ↗ - M3.1 Flash · 公开权重文件核对
Fixed54e4e69d card,48 main shards/DSpark, NVFP4 serving and community license read. This is the check date, not a weight-release date inferred from repository timestamps.
Original release ↗ - MiniMax H3 · 音画同生开放权重核对
Model card and license read: the H3-Omni-Transformer is a 33B dense single-stream Transformer (~13B AdaLN not loaded at inference), 4–15 s at 24 fps 32 kHz stereo up to 2K; the HF repo was created July 28 (API-checked Nov 9), the license text states release/license date 2026-08-02, and the applicable territory excludes the US, EU, UK and South Korea; the community compression derivative (HyperFlow H3 Flash) ships under the same license and is not an official product.
Original release ↗