---
title: Minimax-h3-Turbo-SLA
canonical_url: "https://www.modelscope.cn/models/lightx2v/Minimax-h3-Turbo-SLA"
md_url: "https://www.modelscope.cn/models/lightx2v/Minimax-h3-Turbo-SLA.md"
repository: lightx2v/Minimax-h3-Turbo-SLA
last_updated: 2026-08-20
license: apache-2.0
pipeline_tag: image-to-video
tasks:
  - image-to-video
base_model:
  - MiniMaxAI/MiniMax-H3
base_model_relation: adapter
parameters: 1.7B
tensor_type:
  - BF16
  - F32
library_name:
  - lora
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
downloads: 763
stars: 6
tags:
  - video-generation
  - image-to-video
  - audio-video-generation
  - lora
  - distillation
  - sparse-attention
  - sla
  - lightx2v
---

# Minimax-h3-Turbo-SLA

> Minimax-h3-Turbo-SLA - lightx2v 在 ModelScope 开源的模型。MiniMax-H3 Turbo-SLA

lightx2v/Minimax-h3-Turbo-SLA 是 ModelScope 魔搭社区上的 1.7B 参数image-to-video模型，采用 apache-2.0 许可，基于 MiniMaxAI/MiniMax-H3 构建。

- **Repository**: lightx2v/Minimax-h3-Turbo-SLA
- **License**: apache-2.0
- **Tasks**: image-to-video
- **Parameters**: 1.7B
- **Base model**: MiniMaxAI/MiniMax-H3
- **Tags**: video-generation, image-to-video, audio-video-generation, lora, distillation, sparse-attention, sla, lightx2v
- **Downloads**: 763
- **Stars**: 6
- **Last updated**: 2026-08-20

Source: https://www.modelscope.cn/models/lightx2v/Minimax-h3-Turbo-SLA

---

# MiniMax-H3 Turbo-SLA

MiniMax-H3 Turbo-SLA is a 4-step distilled FL2V checkpoint for [MiniMax-H3](https://huggingface.co/MiniMaxAI/MiniMax-H3), enhanced with [SLA (Sparse–Linear Attention)](https://github.com/thu-ml/SLA) for more efficient inference.

The model uses an **85% attention sparsity ratio**. In our tested [LightX2V](https://github.com/ModelTC/LightX2V) setup, it delivers approximately **2.5× inference acceleration on an NVIDIA RTX 5090** while retaining competitive visual quality.

> This repository contains LoRA weights and requires the original MiniMax-H3 model for inference. Actual performance may vary with resolution, video length, software environment, and hardware configuration.

## Demo

Side-by-side comparison between the 30-step MiniMax-H3 base model and the 4-step MiniMax-H3 Turbo-SLA model:

<video controls loop muted playsinline width="100%">
  <source src="https://cdn-uploads.huggingface.co/production/uploads/680de13385293771bc57400b/HXV6GbED84PeLteFXqK5l.mp4" type="video/mp4">
  Your browser does not support the video tag.
</video>

## Model Highlights

- **4-step distillation** for substantially reduced denoising steps.
- **SLA sparse attention** with an **85% sparsity ratio**.
- Approximately **2.5× inference acceleration on RTX 5090** in our tested LightX2V setup.
- **768p FL2V** generation support.
- Native LightX2V and ComfyUI-compatible BF16 checkpoints.

## Available Checkpoints

| Checkpoint | Format | Description |
|---|---|---|
| [`minimax_h3_fl2v_turbo_4step_v0.1_768p_sla_bf16.safetensors`](https://huggingface.co/lightx2v/Minimax-h3-Turbo-SLA/blob/main/minimax_h3_fl2v_turbo_4step_v0.1_768p_sla_bf16.safetensors) | LightX2V | Native BF16 LoRA checkpoint for LightX2V inference. |
| [`minimax_h3_fl2v_turbo_4step_v0.1_768p_sla_comfyui_bf16.safetensors`](https://huggingface.co/lightx2v/Minimax-h3-Turbo-SLA/blob/main/minimax_h3_fl2v_turbo_4step_v0.1_768p_sla_comfyui_bf16.safetensors) | ComfyUI | Converted BF16 LoRA checkpoint for ComfyUI workflows. |

## LightX2V Inference

For installation and general MiniMax-H3 inference instructions, refer to the [LightX2V MiniMax-H3 examples](https://github.com/ModelTC/LightX2V/tree/main/examples/minimax_h3).

Use the following configuration for SLA-enabled inference on an RTX 5090:

[`minimax_h3_fp8_4step_5090_with_fp8_vae_sla.json`](https://github.com/ModelTC/LightX2V/blob/main/configs/minimax_h3/dmd/minimax_h3_fp8_4step_5090_with_fp8_vae_sla.json)

Key SLA-related settings in this configuration include:

```json
{
  "attn_type": "dynamic_sparse_attn",
  "dynamic_sparse_attn_setting": {
    "sparsity_ratio": 0.85,
    "operator": "sage2"
  },
  "video_flow_shift": 6.0,
  "audio_flow_shift": 3.0,
  "h3_step_update": "training_euler"
}
```

Use the configuration file linked above as the source of truth for the complete inference setup, including FP8 DiT/VAE options and checkpoint paths.

## ComfyUI

For ComfyUI workflows, use:

[`minimax_h3_fl2v_turbo_4step_v0.1_768p_sla_comfyui_bf16.safetensors`](https://huggingface.co/lightx2v/Minimax-h3-Turbo-SLA/blob/main/minimax_h3_fl2v_turbo_4step_v0.1_768p_sla_comfyui_bf16.safetensors)

The ComfyUI checkpoint is a converted version of the SLA-enabled 4-step LoRA and is intended for compatible MiniMax-H3 ComfyUI workflows.

## Related Projects

- [MiniMax-H3](https://huggingface.co/MiniMaxAI/MiniMax-H3)
- [MiniMax-H3 Turbo](https://huggingface.co/lightx2v/Minimax-h3-Turbo)
- [LightX2V](https://github.com/ModelTC/LightX2V)
- [SLA: Sparse–Linear Attention](https://github.com/thu-ml/SLA)

## Acknowledgements

This work builds on MiniMax-H3, LightX2V, and SLA. We thank the respective authors and contributors for making their work available to the community.

## Citation

If you use SLA in your work, please cite:

```bibtex
@article{zhang2025sla,
  title={SLA: Beyond Sparsity in Diffusion Transformers via Fine-Tunable Sparse-Linear Attention},
  author={Zhang, Jintao and Wang, Haoxu and Jiang, Kai and Yang, Shuo and Zheng, Kaiwen and Xi, Haocheng and Wang, Ziteng and Zhu, Hongzhou and Zhao, Min and Stoica, Ion and others},
  journal={arXiv preprint arXiv:2509.24006},
  year={2025}
}
```

## License

The adapter weights in this repository are released under the Apache 2.0 License. Use of the MiniMax-H3 base model is also subject to its corresponding license and terms of use.