Minimax H3 new comfy Model
retained text
In official comfy huggingface there's a new interesting model (w6a8), it seems perfect for 16gb VRAM, i wanted to tried it but Comfy gives error, has someone successfuly tried it? [https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main/diffusion\_models](https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main/diffusion_models) EDIT: i think we need to wait for an update because that quantization is not yet supported
document fetched 2026-09-30 00:35 · canonical https://reddit.com/r/StableDiffusion/comments/1wtgm1i/minimax_h3_new_comfy_model