minimax-h3-nvfp4-fp8

rockerBOO/minimax-h3-nvfp4-fp8

minimax-h3-nvfp4-fp8 is a text-to-video model from rockerBOO. Use it when you need video from a text prompt. It is set up for minimax-h3. The card lists the license as other.

A single ComfyUI checkpoint for MiniMax H3: NVFP4 for the bulk MLP weights, FP8 (tensor-scale) for attn.qkvproj, AdaLN modulation pruned to a lookup table. fl2va (first/last-frame) mode only. Requires a Blackwell GPU…

Downloads · 30 days

224

1% of all-time downloads

All-time downloads

34.1K

Public

Repo size

20.1 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors20.1 GB · 100%

At a glance

Task: Text-to-Video
Library: minimax-h3
License: other
Access: Public
Created: Aug 21, 2026
Updated: Aug 21, 2026
SHA: 86071492

Base models

MiniMaxAI/MiniMax-H3

Model card

From the Hugging Face model README

MiniMax H3 — NVFP4 + FP8 (fl2va)

A single ComfyUI checkpoint for MiniMax H3: NVFP4 for the bulk MLP weights, FP8 (tensor-scale) for attn.qkv_proj, AdaLN modulation pruned to a lookup table. fl2va (first/last-frame) mode only. Requires a Blackwell GPU (RTX 50-series, B100/B200).

File	Size	What it is
`minimax_h3_fl2va_pruned_nvfp4_fp8.safetensors`	20 GB	NVFP4 MLP + FP8 `attn.qkv_proj`, AdaLN-pruned

attn.qkv_proj is quantized fresh from the original BF16 weights (default ctq FP8 output) and spliced into the NVFP4 base. Tested and confirmed working in ComfyUI.

Looking for other variants?

See rockerBOO/minimax-h3-nvfp4-convrot for the full set — ref2va support, the recommended INT8 ConvRot attn.qkv_proj variant (same size, faster), the unpruned base, experimental INT4 ConvRot variants, and full quantization-method documentation. This file is the FP8 speed-comparison baseline for that repo's ConvRot INT8 variant.

License

MiniMax H3 Community License Agreement, inherited from MiniMaxAI/MiniMax-H3 (repackaged by Comfy-Org/MiniMax-H3).

Task: Text-to-Video
Library: minimax-h3
License: other
Created: Aug 21, 2026
Updated: Aug 21, 2026