# MiniMax H3

> MiniMax's H3 video model for high-quality text-to-video, first and last frame, and multimodal reference clips up to 2K with native audio.

Agent resources: the index of everything machine-readable is https://www.ohmygpt.com/llms.txt; the integration skill (endpoints, auth, SDK and coding-tool setup) is https://www.ohmygpt.com/skill.md.

- Model ID: `minimax/minimax-h3`
- Status: available
- Released: 2026-07-31
- Context: 4K tokens
- Input: text, image, video, audio; output: video

## Prices (USD)

| Item | Price | Unit |
| --- | --- | --- |
| Video | $0.073 | per second |
| Image input | $0.0292 | per image |

## Providers

Without a prefix each request goes to the best-priced healthy provider. To pin one, send the model as `<provider>:minimax/minimax-h3`.

| Provider | Pinned model ID | Location | Input | Output | Trains on your data |
| --- | --- | --- | --- | --- | --- |
| MiniMax | `minimax:minimax/minimax-h3` | China | — | — | no |

Web page: https://www.ohmygpt.com/models/minimax/minimax-h3