Home / File reference / INT8 ConvRot FastH3 VSA 4-step

MiniMax H3 INT8 ConvRot FastH3 VSA 4-step: Compatible GPUs, VRAM and Download

Main model (diffusion) · INT8 ConvRot · 22.9GB · updated 2026-09-18

Item	Value
File name	`minimax_h3_fastvideo_vsa_datafree_1300step_4step_int8_convrot.safetensors (Kijai ComfyUI conversion)`
Component	Main model (diffusion)
Quant / format	INT8 ConvRot
Size	22.9 GB
ComfyUI folder	diffusion_models (main model folder)
Source	FastVideo（GitHub hao-ai-lab/FastVideo）· Kijai/MiniMax-H3-experimental

FastVideo FastH3 Preview v1 (with Nuva Lab and NVIDIA FastGen): a 4-step distilled model (step-1300, T2AV) trained with data-free DMD2 + VSA (Video Sparse Attention, arXiv:2505.13389) at 90% sparsity and tile 64. The officially validated route is 4×B200 + fastvideo-kernel; for the LoRA form see FastVideo-FastH3-4-step-Preview-v1-LoRA. This is Kijai's int8 experimental ComfyUI conversion.

Which GPUs can run it

736 × 416 · 5s: 40 GPUs workable

GTX 1060 6GB GTX 1070 8GB GTX 1080 8GB GTX 1080 Ti 11GB RTX 2060 6GB RTX 2060 12GB RTX 2060 Super 8GB RTX 2070 Super 8GB RTX 2080 Ti 11GB RTX 3050 8GB RTX 3060 8GB RTX 3060 12GB RTX 3060 Ti 8GB RTX 3070 8GB RTX 3070 Ti 8GB RTX 3080 10GB RTX 3080 12GB RTX 3080 Ti 12GB +22 more

1344 × 768 · 5s: 40 GPUs workable

1344 × 768 · 15s: 29 GPUs workable

RTX 3060 12GB RTX 3060 Ti 8GB RTX 3070 8GB RTX 3070 Ti 8GB RTX 3080 10GB RTX 3080 12GB RTX 3080 Ti 12GB RTX 3090 24GB RTX 4060 8GB RTX 4060 Ti 8GB RTX 4060 Ti 16GB RTX 4070 12GB RTX 4070 Super 12GB RTX 4070 Ti 12GB RTX 4070 Ti Super 16GB RTX 4080 16GB RTX 4080 Super 16GB RTX 4090 24GB +11 more

“Workable” = estimated status ok or warn with 32GB RAM / fast-disk (warn means weight offloading and noticeably slower).

Download · HuggingFace →

→ Run the numbers for your RAM/budget in the calculator (this file pre-filled)

FAQ

How much VRAM does minimax_h3_fastvideo_vsa_datafree_1300step_4step_int8_convrot.safetensors (Kijai ComfyUI conversion) need?

The file is 22.9GB (Main model (diffusion)). Runtime VRAM also depends on resolution and duration; a 720p-class short clip typically needs roughly 14GB+ of free VRAM. Run the calculator for your exact card.

Where does this file go in ComfyUI?

diffusion_models (main model folder). Keep the filename as-is; refresh ComfyUI and select it in the matching loader.

How does it compare with other quantizations?

See sibling entries in the model library and the INT8/BF16 comparison article. Rule of thumb: 8GB → GGUF Q3/Q2; 12-16GB → pruned INT8 or W4A8; 24GB+ → full INT8 or NVFP4 (RTX 50).

Similar files: BF16 · INT8 · P- BF16 · P- INT8 · P- FP8 · BF16 · INT8 · P- BF16 · P- INT8 · P- FP8