MiniMax H3
Key facts · verified Oct 14, 2026
MiniMax H3 is a MiniMax video generation model released on 31 Jul 2026. Pricing: $0.08 per second (768P), $0.13 per second (2K) on the MiniMax API; open weights under the MiniMax H3 Community License. It ranks #8 on LMArena's text-to-video leaderboard (Elo 1460, read 2 Nov 2026). Source: huggingface.co.
MiniMax's H3, an open multimodal video model that takes text, image, video and audio inputs and generates 4- to 15-second clips at 768P or 2K through the MiniMax API, with synchronised stereo audio. The open release (H3-Base) generates 768p video at 24 fps; its licence lists the United States, European Union, United Kingdom and South Korea as excluded territories that need a separate licence.
$0.08 per second (768P), $0.13 per second (2K) on the MiniMax API; open weights under the MiniMax H3 Community License
Not published
Key Features
Full Specifications
$0.08 per second (768P), $0.13 per second (2K) on the MiniMax API; open weights under the MiniMax H3 Community License | |
|---|---|
| Identity | |
| Developer | MiniMax |
| Released | Jul 2026 |
| Licence | MiniMax H3 Community License — US, EU, UK and South Korea excluded without a separate licence |
| Self-hostable | Yes — H3-Base weights (768p); 2K and Context-IR are platform-only |
| Cost | |
| Price per second | $0.08 (768P), $0.13 (2K) |
| Capacity | |
| Max duration | 15 seconds (4 to 15) |
| Max resolution | 2K via the API; 768p self-hosted |
| FPS | 24 |
| Capability | |
| Text-to-video | Yes |
| Image-to-video | Yes — first and last frame |
| Native audio | Yes — synchronised stereo audio |
| Camera controls | Reference images, video or audio for character, motion, camera, style or voice |
| Access | |
| API | Yes — MiniMax API |
| Measured quality | |
| LMArena Elo | 1460 (checked Nov 2026) |
| LMArena rank | Rank 8 |