endman100/MiniMax-H3-Ref2VA-FP8

MiniMax H3 Ref2VA FP8 E4M3FN for ComfyUI

Unofficial mixed-precision FP8 conversion of the MiniMax H3 Ref2VA diffusion model for ComfyUI.

Files

• minimax_h3_ref2va_fp8_e4m3fn.safetensors
• Size: 33,236,581,696 bytes (30.954 GiB)
• SHA-256: EE41F81E1347B415000F5818712E5AACD3C841851AF56F66FC907F2CCBA43151

Conversion

The source BF16 checkpoint was converted tensor-by-tensor to avoid loading the full model into RAM or VRAM. Multi-dimensional weight tensors were converted to FP8 E4M3FN, while normalization, bias, conditioning, final-layer, and time-embedding tensors retained BF16 or FP32 precision.

• FP8 E4M3FN tensors: 258
• BF16 tensors: 264
• FP32 tensors: 13
• Source SHA-256: E32C54C1A7B4F5F397F195CEA267CCB18806303BB665678C4BEE60953BDF3026

This is a modified derivative and is not an official MiniMax release.

ComfyUI

Place the checkpoint in ComfyUI/models/diffusion_models/ and load it with the standard UNETLoader using weight_dtype: default.

Verified locally with ComfyUI 0.29.0 and an NVIDIA RTX 5090 using:

• Ref2VA
• 768 x 1344
• 124 frames at 24 fps
• 20 steps
• res_multistep sampler
• beta scheduler

ComfyUI reported the checkpoint weight dtype as torch.float8_e4m3fn with BF16 manual casting during inference.

License and source

Base model: MiniMaxAI/MiniMax-H3

This repository includes the upstream MiniMax H3 Community License Agreement and required NOTICE. The upstream license contains territorial restrictions and excludes the European Union, United Kingdom, Republic of Korea, and United States from the applicable territory unless separately authorized by MiniMax. Review the full LICENSE before downloading, using, or redistributing this checkpoint.