MiniMax-H3 supports text-to-video, image-to-video (first frame / last frame / first and last frames), and multimodal reference-to-video generation, with 768P and 2K resolution output.
Inference Service Provider
The inference service provider for the MiniMax/MiniMax-H3 model is MiniMax.
Model Capabilities
| Capability | Support | Capability | Support |
|---|---|---|---|
Input modality | Image Text Video Audio | Output modality | Video |
Model playground | Unsupported | Function Calling | Unsupported |
Structured output | Unsupported | Web search | Unsupported |
Prefix completion | Unsupported | Context caching | Unsupported |
Batch inference | Unsupported | Model fine-tuning | Unsupported |
Context Limits
| Parameter | Value | Parameter | Value |
|---|---|---|---|
Maximum input length | - | Maximum output length | - |
Context length | - |
Pricing
This section only displays the original prices for model calls and does not include limited-time promotions or other offers. Please visit the Model Studio console for promotional pricing.
- China (Beijing)
| Model/API | Resolution | Billing Rule | List Price |
|---|---|---|---|
MiniMax-H3 | 2K | Per second | CNY 0.80/second |
MiniMax-H3 | 768P | Per second | CNY 0.50/second |
| Model/API | Material Type | Billing Rule |
|---|---|---|
MiniMax-H3 | Audio | Free |
MiniMax-H3 | Image | Free for up to 5 images; excess charged at CNY 0.20/image |
MiniMax-H3 | Video | Billed by input video duration and output video resolution: 2K CNY 0.80/second, 768P CNY 0.50/second |
Rate Limits
- China (Beijing)
| Parameter | Value |
|---|---|
RPM (Requests per minute) | 300 |