Oriel — AI & technology worth noticing, every day.

Abiray/MiniMax-H3-Image-GGUF

AI model · https://huggingface.co/Abiray/MiniMax-H3-Image-GGUF

First seen by Oriel on .

AI summary

This repository provides GGUF quantizations of the 33.1B MiniMax-H3-Image pruned DiT, preserving sensitive layers and supporting native stock ComfyUI-GGUF for six-step generation.

  • Five quantizations range from 6.88 GB Q2_K to 14.10 GB Q5_K_M, targeting approximately 8–24 GB of VRAM.
  • The input patchifier, PDD output head, and AdaLN modulation basis remain unquantized in FP32 to reduce quantization noise and grid artifacts.
  • Usage requires the dedicated 24-channel Image VAE, a Qwen3-VL-32B text encoder, ComfyUI v0.45.0+, and ComfyUI-GGUF.
  • Recommended inference uses Euler with six steps, CFG 1.0, denoise 1.0, and resolutions from 1024 × 1024 up to 4096 × 4096.
  • The model is available for research, personal, and commercial creation under the MiniMax-H3-Image Community License Agreement, subject to stated conditions.

#gguf #comfyui #text-to-image #diffusion transformers

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Likes

Total likes reported by the source.

  • Hugging Face Models: 291, latest observation ; +110 over observed period since

Downloads (30d)

Rolling 30-day downloads. A change in this value is not new downloads during the observed period.

  • Hugging Face Models: 155,475, latest observation ; +103,227 over observed period since

Aliases

SourceExternal IDURL
Hugging Face ModelsAbiray/MiniMax-H3-Image-GGUFhttps://huggingface.co/Abiray/MiniMax-H3-Image-GGUF Read discussion