Only models that load directly in ComfyUI, screened by downloads, likes and community reputation (data from HuggingFace, 2026-09-18). Original Diffusers format, MLX, Wan2GP and other runtimes are not listed. ★ = editor’s pick. Downloads go through a short interstitial page (an ad keeps this site free) before opening the target link.
★ Editor’s pick
ComfyUI official repack (Comfy-Org)
Comfy-Org/MiniMax-H3
All files used by this site's calculator come from here: bf16 / int8_convrot / pruned / fp8_scaled + TE + VAE + turbo LoRA + ControlNet pack
⬇ downloads 20.0M ♥ likes 2k size 343GB (all variants)
official ComfyUI must-have
Download · HuggingFace Download · ModelScope
★ Editor’s pick
GGUF full version (Abiray, unpruned)
Abiray/MiniMax-H3-GGUF
Unpruned GGUF, Q2-Q8 tiers, loads directly with ComfyUI-GGUF nodes; even 8GB VRAM can give it a try
⬇ downloads 1.0M ♥ likes 134 size Q2-Q8 tiers
GGUF ComfyUI
★ Editor’s pick
Pruned GGUF (Abiray pruned version)
Abiray/MiniMax-H3-Pruned-GGUF
Pruned + GGUF, Q3_K_M is only 8.9GB (data source for this site's VRAM-saving tier)
⬇ downloads 143k ♥ likes 65 size 8.9-21.6GB
GGUF pruned ComfyUI
★ Editor’s pick
Fused Turbo INT8 (MATLOWAI, pre-fused acceleration)
MATLOWAI/minimax-h3-fused-turbo-int8-convrot
Turbo LoRA and Ref2VA delta fused straight into INT8 weights: one less LoRA load, more stable VRAM, ready for 4-8 steps out of the box
⬇ downloads 512k ♥ likes 198 size single file (fused)
INT8 fused hassle-free ComfyUI
★ Editor’s pick
larryvrh Turbo LoRA (v4, hottest community accelerator)
larryvrh/MiniMax-H3-Turbo-Lora
8-step distillation, the community's best-reviewed general accelerator; use with the dual-clock sampler
⬇ downloads 218k ♥ likes 989 size ≈1.8GB
8-step LoRA ComfyUI
★ Editor’s pick
Turbo LoRA ComfyUI conversion + low-rank compression (drbaph)
drbaph/MiniMax-H3-Turbo-Lora-ComfyUI
ComfyUI-compatible conversion of larry/lightx2v, with dynamic-rank compressed variants (rank28 is only 0.37GB)
⬇ downloads 197k ♥ likes 441 size 0.37-0.9GB
4/8-step LoRA ComfyUI
★ Editor’s pick
Kijai experimental repo (W4A8 weights / int8 VAE / VSA distillation / PDD)
Kijai/MiniMax-H3-experimental
The only source for this site's W4A8 (12.5GB), 1.5x int8 VAE, FastH3 VSA int8, and the PDD comfy conversion
⬇ downloads — ♥ likes 408 size per file
W4A8 experimental ComfyUI
★ Editor’s pick
Photorealistic people LoRA (fal)
fal/MiniMax-H3-Realism-People-LoRA
Realistic/cinematic people finetune; README explicitly supports ComfyUI (Load LoRA works as-is, no conversion), trigger word r34l1sm
⬇ downloads 61k ♥ likes 400 size LoRA（rank32）
realism portrait ComfyUI-verified
★ Editor’s pick
Text Encoder GGUF (16GB RAM lifesaver)
joeygambino/MiniMax-H3-encoder-GGUF
GGUF build of the Qwen3-VL-32B text encoder, loaded with ComfyUI-GGUF, easing encoder residency issues on low-RAM machines
⬇ downloads 216k ♥ likes 12 size ~13GB
TE GGUF low RAM ComfyUI
NVFP4 / INT4 / INT8 quant collection (Abiray)
Abiray/Minimax-H3-nvfp4-INT4-INT8-Convrot
Multi-format quantization collection (NVFP4 needs 50-series); good for quick comparison
⬇ downloads 210k ♥ likes 232 size per file
NVFP4 INT4 INT8 ComfyUI
Ref-Delta fused (xmarre): FL2VA+Ref2VA single file
xmarre/MiniMax-H3-Pruned-Ref-Delta-Fused-r1024-ComfyUI
Ref2VA delta fused into pruned INT8; one file covers both modes and saves disk space and load switching
⬇ downloads 95k ♥ likes 31 size single file
INT8 fused experimental ComfyUI
W4A8 ConvRot (AX1Y2JP)
AX1Y2JP/MiniMax-H3-W4A8-ConvRot
Community W4A8 (4-bit weights + INT8 compute) conversion; requires ComfyUI ≥0.31
⬇ downloads 35k ♥ likes 27 size ≈12GB
W4A8 ComfyUI
NVFP4 ConvRot (rockerBOO)
rockerBOO/minimax-h3-nvfp4-convrot
Early NVFP4 conversion (Blackwell), ~12GB
⬇ downloads 18k ♥ likes 28 size ≈12GB
NVFP4 50-series ComfyUI
Alibaba PAI official PDD acceleration (8-step / 4-step)
alibaba-pai/MiniMax-H3-Acc-LoRAs
Parallel Decoding Distillation, CFG-free; requires the ComfyUI PDD node pack (the nodes can read the official files directly) and cannot stack with other distillation
⬇ downloads 89k ♥ likes 235 size 1.37GB
8-step official distillation PDD ComfyUI nodes
PDD ComfyUI conversion (aptech0081)
aptech0081/MiniMax-H3-Acc-LoRAs-ComfyUI
ComfyUI key conversion of the alibaba-pai PDD weights (for use with the PDD node pack)
⬇ downloads 45k ♥ likes 71 size 1.37GB
PDD ComfyUI
FastH3 LoRA + VSA nodes (ComfyUI-ready)
barelymining/ComfyUI-MiniMax-H3-FastVideo
ComfyUI-loadable LoRA for the FastVideo 4-step distillation, plus companion nodes that route attention to the VSA sparse path (with the 4-step distillation they give minute-level drafts on 24GB)
⬇ downloads 17k ♥ likes 15 size per file
4-step LoRA ComfyUI
Singularity finetune (HDR / quality boost)
WarmBloodAban/Minimax-h3_Singularity
Highly upvoted all-round community finetune with an HDR/texture lean; check its usage terms and license
⬇ downloads 182k ♥ likes 484 size per file
finetune HDR ComfyUI
Physics enhancement LoRA (Jojocodex)
Jojocodex/minimax-h3-spatial-physics-lora
Boosts physical realism — collisions, gravity, stacking — at strength 0.3-0.5 and stacks with turbo; trained on Comfy-Org weights and works with the built-in loader
⬇ downloads 25k ♥ likes 116 size 148MB
physics stackable ComfyUI
Camera motion LoRA (Jojocodex)
Jojocodex/minimax-h3-Camera-Motion-lora
Stable camera work and push/pull/pan/tilt control, good for strengthening shot language; trained on Comfy-Org weights
⬇ downloads 17k ♥ likes 69 size LoRA
camera motion ComfyUI
Fun ControlNet Union (ComfyUI repack)
Comfy-Org/MiniMax-H3 · model_patches/
Pose/depth/edge control nets (H3-Fun series); Comfy-Org has packed them in native ComfyUI format (model_patches/), the original Diffusers version is not listed here
⬇ downloads — ♥ likes — size 2.3-4.2GB
ControlNet ComfyUI