Curated download URLs and target directories

Model registry

by artokun · MIT license · ★ 759 on GitHub · updated Sep 21

Curated download URLs and target directories, organized by family (Flux, WAN, LTX, Qwen, Z-Image, SD15/SDXL), for every model the comfyui-mcp skills reference, covering checkpoints, VAEs, text encoders, and LoRAs. Use when downloading…

---
name: model-registry
description: Curated download URLs and target directories, organized by family (Flux, WAN, LTX, Qwen, Z-Image, SD15/SDXL), for every model the comfyui-mcp skills reference, covering checkpoints, VAEs, text encoders, and LoRAs. Use when downloading models with download_model (action:"download" / action:"download_civitai"), when a workflow fails with a missing-model error, or when setting up a new machine.
---

# Model Registry

One table per family: filename → source URL → target subdir under
`<COMFYUI>/models/`. Use with `download_model({ action: "download", url, target_subfolder, filename })`.
This registry grows with every release. If a model you need is missing, use
`action:"search"` (HuggingFace) or `action:"download_civitai"` and consider
contributing the row.

**Quant column (kitchen).** Filenames in this registry are mostly bf16 / fp8. NVFP4 and MXFP8 siblings, when a vendor publishes them, are the same stem with `nvfp4` / `mxfp8` in the name and belong in `diffusion_models/` too. `kitchen` action:"assess" only offers an NVFP4 swap when that sibling is **already listed locally** — it does not guess a URL from the filename. Blackwell (SM ≥ 10.0) is the GPU that can run NVFP4 / MXFP8 compute.

**Conventions**
- HF "resolve" URLs download directly: `https://huggingface.co/<repo>/resolve/main/<path>`
- 🔒 = gated repo, needs `HUGGINGFACE_TOKEN` (accept the license on the HF page first)

Overview

One table per family: filename → source URL → target subdir under <COMFYUI>/models/. Use with download_model({ action: "download", url, target_subfolder, filename }). This registry grows with every release. If a model you need is missing, use action:"search" (HuggingFace) or action:"download_civitai" and consider contributing the row.

Manual install (npx, ZIP)

Claude Codenpx skills add artokun/comfyui-mcp --skill model-registry -g -a claude-code

Claude Code (manual)

mkdir -p ~/.claude/skills && curl -sL https://codeload.github.com/artokun/comfyui-mcp/tar.gz/1d5b12b1b499 \
  | tar -xz -C ~/.claude/skills --strip-components=3 comfyui-mcp-1d5b12b1b499/plugin/skills/model-registry

Windows (PowerShell)

New-Item -ItemType Directory -Force "$HOME\.claude\skills" | Out-Null; Invoke-WebRequest "https://m.agentalley.io/zip/f1fd507cb3cea01e547a/model-registry.zip" -OutFile "$env:TEMP\model-registry.zip"; Expand-Archive "$env:TEMP\model-registry.zip" -DestinationPath "$HOME\.claude\skills" -Force

Claude.ai / desktopDownload the ZIP and upload it under Settings → Capabilities → Skills.

Codex, Cursor, other agentsPoint them at SKILL.md. The other files are plain text.

Connects with Agent Alley MCP · agentalley.io/mcp

Files included

SKILL.md8.2 KB

Read the files here

artokun/1d5b12b1 file

SKILL.md8.2 KB

SKILL.md

Show the full text125 lines

model-registry/SKILL.md125 lines · 8.2 KB

PreviewCode

Outline

RawView on GitHub

Model Registry

One table per family: filename → source URL → target subdir under <COMFYUI>/models/. Use with download_model({ action: "download", url, target_subfolder, filename }). This registry grows with every release. If a model you need is missing, use action:"search" (HuggingFace) or action:"download_civitai" and consider contributing the row.

Quant column (kitchen). Filenames in this registry are mostly bf16 / fp8. NVFP4 and MXFP8 siblings, when a vendor publishes them, are the same stem with nvfp4 / mxfp8 in the name and belong in diffusion_models/ too. kitchen action:"assess" only offers an NVFP4 swap when that sibling is already listed locally — it does not guess a URL from the filename. Blackwell (SM ≥ 10.0) is the GPU that can run NVFP4 / MXFP8 compute.

Conventions

HF "resolve" URLs download directly: https://huggingface.co/<repo>/resolve/main/<path>
🔒 = gated repo, needs HUGGINGFACE_TOKEN (accept the license on the HF page first)
CivitAI model-page URLs need download_model action:"download_civitai" (resolves version → file); raw civitai.com/api/download/... URLs work with action:"download" + CIVITAI_API_TOKEN
Always verify the exact filename a workflow's loader expects. The model-compatibility skill covers which VAE/CLIP pairs with which architecture

Shared VAEs & text encoders (download these once)

File	Source	Target
`ae.safetensors` (Flux/Z-Image VAE)	`huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors` (Apache, not gated)	`vae/`
`vae-ft-mse-840000-ema-pruned.safetensors` (SD1.5 VAE)	`huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors`	`vae/`
`clip_l.safetensors`	`huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors`	`text_encoders/`
`t5xxl_fp8_e4m3fn.safetensors`	`huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors`	`text_encoders/`
`umt5_xxl_fp8_e4m3fn_scaled.safetensors` (WAN)	`huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors`	`text_encoders/`

Flux (see `flux-txt2img` skill for workflows)

File	Source	Target
`flux1-dev.safetensors` 🔒	`huggingface.co/black-forest-labs/FLUX.1-dev`	`diffusion_models/`
`flux1-schnell.safetensors`	`huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors`	`diffusion_models/`
`flux2-vae.safetensors` (Flux 2 Klein)	`huggingface.co/Comfy-Org/flux2-klein` — check repo for current path	`vae/`

WAN 2.x video (see `wan-t2v-video` / `wan-flf-video` skills)

Comfy-Org repackages everything: huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged under split_files/.

File	Path in repo	Target
`wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors`	`split_files/diffusion_models/`	`diffusion_models/`
`wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors`	`split_files/diffusion_models/`	`diffusion_models/`
`wan_2.1_vae.safetensors`	`split_files/vae/`	`vae/`
`wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors`	`split_files/loras/`	`loras/`
`wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors`	`split_files/loras/`	`loras/`
Magical-morph I2V LoRAs (high/low)	`huggingface.co/NikolaSigmoid/wan2.2-i2v-loras-magical-morph`	`loras/`
SkinMorph Redmond I2V 14B	`civitai.com/models/2210162` (use `action:"download_civitai"`)	`loras/`

LTX-2 19B video (see `ltxv2-video` skill)

File	Source	Target
`ltx-2-19b-distilled.safetensors`	`huggingface.co/Lightricks/LTX-2` — check repo for the bf16 distilled path	`checkpoints/`
`gemma_3_12B_it_fp4_mixed.safetensors`	`huggingface.co/Comfy-Org` LTX repackage	`text_encoders/`
`ltx-2-spatial-upscaler-x2-1.0.safetensors`	`huggingface.co/Lightricks/LTX-2`	`upscale_models/`
Camera-control LoRAs (`dolly-left`, distilled-384)	`huggingface.co/Lightricks/LTX-2` LoRA dir	`loras/`

Qwen-Image (see `qwen-txt2img` / `qwen-image-edit` skills)

File	Source	Target
`qwen_image_fp8_e4m3fn.safetensors`	`huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors`	`diffusion_models/`
`qwen_2.5_vl_7b_fp8_scaled.safetensors`	same repo, `split_files/text_encoders/`	`text_encoders/`
`qwen_image_vae.safetensors`	same repo, `split_files/vae/`	`vae/`
`qwen_image_edit_2511_bf16.safetensors`	`huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI` — check repo	`diffusion_models/`
Qwen-Image 2512 Lightning LoRA	`huggingface.co/lightx2v/Qwen-Image-2512-Lightning`	`loras/`
Qwen-Image-Edit 2511 Lightning LoRA	`huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning`	`loras/`
Qwen-Image 2512 Turbo LoRA	`huggingface.co/Wuli-art/Qwen-Image-2512-Turbo-LoRA`	`loras/`

MiniMax H3 video (see `minimax-h3-video` skill)

Comfy-Org INT8 pack: huggingface.co/Comfy-Org/MiniMax-H3. Local weights are under the MiniMax H3 Community License (territory-restricted). Prefer the filenames the Comfy-Org templates already name.

File	Source	Target
`minimax_h3_fl2va_pruned_int8_convrot.safetensors`	`huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_fl2va_pruned_int8_convrot.safetensors`	`diffusion_models/`
`minimax_h3_ref2va_pruned_int8_convrot.safetensors`	same repo, `diffusion_models/`	`diffusion_models/`
`qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors`	same repo, `text_encoders/`	`text_encoders/`
`minimax_h3_video_vae_fp16.safetensors`	same repo, `vae/`	`vae/`
`minimax_h3_audio_vae_fp32.safetensors`	same repo, `vae/`	`vae/`
`minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors`	`huggingface.co/lightx2v/Minimax-h3-Turbo/resolve/main/minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors`	`loras/`
`minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors`	`huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/loras/minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors`	`loras/`

Z-Image (see `z-image-txt2img` skill)

File	Source	Target
`z_image_turbo_bf16.safetensors`	`huggingface.co/Comfy-Org/z_image_turbo` — check repo for split_files path	`diffusion_models/`
`z_image_base_bf16.safetensors`	same family repo	`diffusion_models/`
`qwen_3_4b.safetensors` (Z-Image text encoder)	Comfy-Org repackage, `split_files/text_encoders/`	`text_encoders/`

SDXL / SD1.5 community checkpoints (CivitAI)

Use download_model({ action: "download_civitai", model_id }). It resolves the latest version and handles auth:

Model	CivitAI
RedCraft RedZ ImageDX	`civitai.com/models/958009`
Qwen Ultimate Realism v1.1	`civitai.com/models/2027494`
Copax Timeless	`civitai.com/models/copaxTimeless` (search by name)

Failure modes

404 on an HF resolve URL. The repo restructured. Open the repo page, find the file under "Files", and rebuild the resolve URL.
401/403. Gated repo. Set HUGGINGFACE_TOKEN after accepting the license, or CIVITAI_API_TOKEN for early-access CivitAI files.
Wrong dropdown after download. The file landed in the wrong models/ subdir; check the Target column and model-compatibility.
Downloads resume on retry (HTTP Range), so re-run the same download_model call after a network drop.

Sources

Official: Hugging Face resolve-URL convention https://huggingface.co/<repo>/resolve/main/<path>; kitchen NVFP4/MXFP8 hardware gates in ComfyUI comfy/model_management.py and quant metadata in comfy/quant_ops.py.
Empirical: the curated list and Civitai notes are ours, not a vendor catalog.

1	---
2	name: model-registry
3	description: Curated download URLs and target directories, organized by family (Flux, WAN, LTX, Qwen, Z-Image, SD15/SDXL), for every model the comfyui-mcp skills reference, covering checkpoints, VAEs, text encoders, and LoRAs. Use when downloading models with download_model (action:"download" / action:"download_civitai"), when a workflow fails with a missing-model error, or when setting up a new machine.
4	---
5
6	# Model Registry
7
8	One table per family: filename → source URL → target subdir under
9	`<COMFYUI>/models/`. Use with `download_model({ action: "download", url, target_subfolder, filename })`.
10	This registry grows with every release. If a model you need is missing, use
11	`action:"search"` (HuggingFace) or `action:"download_civitai"` and consider
12	contributing the row.
13
14	Quant column (kitchen). Filenames in this registry are mostly bf16 / fp8. NVFP4 and MXFP8 siblings, when a vendor publishes them, are the same stem with `nvfp4` / `mxfp8` in the name and belong in `diffusion_models/` too. `kitchen` action:"assess" only offers an NVFP4 swap when that sibling is already listed locally — it does not guess a URL from the filename. Blackwell (SM ≥ 10.0) is the GPU that can run NVFP4 / MXFP8 compute.
15
16	Conventions
17	- HF "resolve" URLs download directly: `https://huggingface.co/<repo>/resolve/main/<path>`
18	- 🔒 = gated repo, needs `HUGGINGFACE_TOKEN` (accept the license on the HF page first)
19	- CivitAI model-page URLs need `download_model` `action:"download_civitai"` (resolves version → file); raw `civitai.com/api/download/...` URLs work with `action:"download"` + `CIVITAI_API_TOKEN`
20	- Always verify the exact filename a workflow's loader expects. The `model-compatibility` skill covers which VAE/CLIP pairs with which architecture
21
22	## Shared VAEs & text encoders (download these once)
23
24	\| File \| Source \| Target \|
25	\|---\|---\|---\|
26	\| `ae.safetensors` (Flux/Z-Image VAE) \| `huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/ae.safetensors` (Apache, not gated) \| `vae/` \|
27	\| `vae-ft-mse-840000-ema-pruned.safetensors` (SD1.5 VAE) \| `huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors` \| `vae/` \|
28	\| `clip_l.safetensors` \| `huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors` \| `text_encoders/` \|
29	\| `t5xxl_fp8_e4m3fn.safetensors` \| `huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors` \| `text_encoders/` \|
30	\| `umt5_xxl_fp8_e4m3fn_scaled.safetensors` (WAN) \| `huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors` \| `text_encoders/` \|
31
32	## Flux (see `flux-txt2img` skill for workflows)
33
34	\| File \| Source \| Target \|
35	\|---\|---\|---\|
36	\| `flux1-dev.safetensors` 🔒 \| `huggingface.co/black-forest-labs/FLUX.1-dev` \| `diffusion_models/` \|
37	\| `flux1-schnell.safetensors` \| `huggingface.co/black-forest-labs/FLUX.1-schnell/resolve/main/flux1-schnell.safetensors` \| `diffusion_models/` \|
38	\| `flux2-vae.safetensors` (Flux 2 Klein) \| `huggingface.co/Comfy-Org/flux2-klein` — check repo for current path \| `vae/` \|
39
40	## WAN 2.x video (see `wan-t2v-video` / `wan-flf-video` skills)
41
42	Comfy-Org repackages everything: `huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged` under `split_files/`.
43
44	\| File \| Path in repo \| Target \|
45	\|---\|---\|---\|
46	\| `wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors` \| `split_files/diffusion_models/` \| `diffusion_models/` \|
47	\| `wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors` \| `split_files/diffusion_models/` \| `diffusion_models/` \|
48	\| `wan_2.1_vae.safetensors` \| `split_files/vae/` \| `vae/` \|
49	\| `wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors` \| `split_files/loras/` \| `loras/` \|
50	\| `wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors` \| `split_files/loras/` \| `loras/` \|
51	\| Magical-morph I2V LoRAs (high/low) \| `huggingface.co/NikolaSigmoid/wan2.2-i2v-loras-magical-morph` \| `loras/` \|
52	\| SkinMorph Redmond I2V 14B \| `civitai.com/models/2210162` (use `action:"download_civitai"`) \| `loras/` \|
53
54	## LTX-2 19B video (see `ltxv2-video` skill)
55
56	\| File \| Source \| Target \|
57	\|---\|---\|---\|
58	\| `ltx-2-19b-distilled.safetensors` \| `huggingface.co/Lightricks/LTX-2` — check repo for the bf16 distilled path \| `checkpoints/` \|
59	\| `gemma_3_12B_it_fp4_mixed.safetensors` \| `huggingface.co/Comfy-Org` LTX repackage \| `text_encoders/` \|
60	\| `ltx-2-spatial-upscaler-x2-1.0.safetensors` \| `huggingface.co/Lightricks/LTX-2` \| `upscale_models/` \|
61	\| Camera-control LoRAs (`dolly-left`, distilled-384) \| `huggingface.co/Lightricks/LTX-2` LoRA dir \| `loras/` \|
62
63	## Qwen-Image (see `qwen-txt2img` / `qwen-image-edit` skills)
64
65	\| File \| Source \| Target \|
66	\|---\|---\|---\|
67	\| `qwen_image_fp8_e4m3fn.safetensors` \| `huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_fp8_e4m3fn.safetensors` \| `diffusion_models/` \|
68	\| `qwen_2.5_vl_7b_fp8_scaled.safetensors` \| same repo, `split_files/text_encoders/` \| `text_encoders/` \|
69	\| `qwen_image_vae.safetensors` \| same repo, `split_files/vae/` \| `vae/` \|
70	\| `qwen_image_edit_2511_bf16.safetensors` \| `huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI` — check repo \| `diffusion_models/` \|
71	\| Qwen-Image 2512 Lightning LoRA \| `huggingface.co/lightx2v/Qwen-Image-2512-Lightning` \| `loras/` \|
72	\| Qwen-Image-Edit 2511 Lightning LoRA \| `huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning` \| `loras/` \|
73	\| Qwen-Image 2512 Turbo LoRA \| `huggingface.co/Wuli-art/Qwen-Image-2512-Turbo-LoRA` \| `loras/` \|
74
75	## MiniMax H3 video (see `minimax-h3-video` skill)
76
77	Comfy-Org INT8 pack: `huggingface.co/Comfy-Org/MiniMax-H3`. Local weights are
78	under the MiniMax H3 Community License (territory-restricted). Prefer the
79	filenames the Comfy-Org templates already name.
80
81	\| File \| Source \| Target \|
82	\|---\|---\|---\|
83	\| `minimax_h3_fl2va_pruned_int8_convrot.safetensors` \| `huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_fl2va_pruned_int8_convrot.safetensors` \| `diffusion_models/` \|
84	\| `minimax_h3_ref2va_pruned_int8_convrot.safetensors` \| same repo, `diffusion_models/` \| `diffusion_models/` \|
85	\| `qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors` \| same repo, `text_encoders/` \| `text_encoders/` \|
86	\| `minimax_h3_video_vae_fp16.safetensors` \| same repo, `vae/` \| `vae/` \|
87	\| `minimax_h3_audio_vae_fp32.safetensors` \| same repo, `vae/` \| `vae/` \|
88	\| `minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors` \| `huggingface.co/lightx2v/Minimax-h3-Turbo/resolve/main/minimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors` \| `loras/` \|
89	\| `minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors` \| `huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/loras/minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors` \| `loras/` \|
90
91	## Z-Image (see `z-image-txt2img` skill)
92
93	\| File \| Source \| Target \|
94	\|---\|---\|---\|
95	\| `z_image_turbo_bf16.safetensors` \| `huggingface.co/Comfy-Org/z_image_turbo` — check repo for split_files path \| `diffusion_models/` \|
96	\| `z_image_base_bf16.safetensors` \| same family repo \| `diffusion_models/` \|
97	\| `qwen_3_4b.safetensors` (Z-Image text encoder) \| Comfy-Org repackage, `split_files/text_encoders/` \| `text_encoders/` \|
98
99	## SDXL / SD1.5 community checkpoints (CivitAI)
100
101	Use `download_model({ action: "download_civitai", model_id })`. It resolves the latest version and
102	handles auth:
103
104	\| Model \| CivitAI \|
105	\|---\|---\|
106	\| RedCraft RedZ ImageDX \| `civitai.com/models/958009` \|
107	\| Qwen Ultimate Realism v1.1 \| `civitai.com/models/2027494` \|
108	\| Copax Timeless \| `civitai.com/models/copaxTimeless` (search by name) \|
109
110	## Failure modes
111
112	- 404 on an HF resolve URL. The repo restructured. Open the repo page,
113	find the file under "Files", and rebuild the resolve URL.
114	- 401/403. Gated repo. Set `HUGGINGFACE_TOKEN` after accepting the
115	license, or `CIVITAI_API_TOKEN` for early-access CivitAI files.
116	- Wrong dropdown after download. The file landed in the wrong `models/`
117	subdir; check the Target column and `model-compatibility`.
118	- Downloads resume on retry (HTTP Range), so re-run the same
119	`download_model` call after a network drop.
120
121	## Sources
122
123	- Official: Hugging Face resolve-URL convention https://huggingface.co/<repo>/resolve/main/<path>; kitchen NVFP4/MXFP8 hardware gates in ComfyUI `comfy/model_management.py` and quant metadata in `comfy/quant_ops.py`.
124	- Empirical: the curated list and Civitai notes are ours, not a vendor catalog.
125