Best Open-Weight Image Model for ComfyUI (2026)
Licences, quantization and hardware for FLUX, Stable Diffusion and newer families
Model names, prices and licences change often. Facts below are as printed on each provider page on the date shown. This is a shortlist for testing, not a benchmark ranking, and we did not run image tests for it.
On this page
Which open-weight image model should I run in ComfyUI?
Direct Summary
Choose the open-weight model by licence, hardware and workflow, not by headline quality. ComfyUI lists many supported families, but FLUX.1 schnell is Apache 2.0 while FLUX.1 dev is non-commercial, and GGUF files need a community node, so read the exact model licence first.
Technical Nuance & Operational Realities
Open weights do not mean unrestricted use. Licences differ even inside one model family, and the same name can cover permissive and non-commercial releases. ComfyUI itself is a separate project under GPL-3.0; the licence of the model you load is set by its publisher.
ComfyUI's README describes "a powerful and modular node graph interface" and lists supported image families including Stable Diffusion 1.5, SDXL, SD3.5, Flux.1, Flux.2, Qwen Image, HiDream and Ideogram 4, plus editing families such as Flux Kontext and Flux.2 Klein. Its README also makes a marketing claim of running large models on "as low as 4GB vram + 8GB ram" using weight streaming; the documentation page on system requirements gives no VRAM minimum, so test on your own hardware.
What to check before you install
Licences of the models people ask about
| Model | Licence as printed | Note |
|---|---|---|
| FLUX.1 [schnell] | Apache 2.0 | Permissive |
| FLUX.1 [dev], Fill, Canny, Depth, Kontext [dev], Krea [dev] | FLUX.1 [dev] Non-Commercial License | Commercial path via Black Forest Labs' licensing page |
| FLUX.2 [klein] 4B | Apache 2.0 | Smaller FLUX.2 model |
| FLUX.2 [dev], FLUX.2 [klein] 9B | FLUX Non-Commercial License | Not permissive |
| FLUX 3 Image Klein (Oct 2026) | FLUX Kommunity License for weights | We could not read the licence text; check it before use |
| Stable Diffusion 3.5 Large | Stability AI Community License | Free under $1M annual revenue with registration; above that Stability decides at its discretion |
| Ideogram 4 weights | Ideogram Non-Commercial Model Agreement | Paid licence needed for commercial deployment |
What ComfyUI adds
The node graph lets you chain loaders, text encoders, samplers, conditioning, upscaling and output. The README lists LoRAs, ControlNets, inpainting, outpainting and quantized-model support. Whether a given feature works depends on the model, your ComfyUI version, custom nodes and hardware.
Quantization and hardware
Quantization can lower memory needs but can change quality or speed, and formats are not interchangeable. ComfyUI's documentation states: "ComfyUI does not natively support GGUF format models", and directs users to community custom nodes. BFL's FLUX 3 Image Klein README says an 8 GB GPU can use int8 files in ComfyUI, and its FLUX.2 README says klein 4B fits in about 8 GB. Check each model card before downloading.
At a Glance
Facts are as printed on provider pages read on 2026-10-21.
| If your priority is | Look at | Check first |
|---|---|---|
| Permissive licence | FLUX.1 [schnell], FLUX.2 [klein] 4B | Exact model file and licence |
| Editing and conditioning tools | FLUX.1 tools, Kontext | Non-commercial default |
| Small GPU | Quantized or small models | Matching loader, quality loss |
| Existing SD extensions | Stable Diffusion 3.5 | Revenue threshold and registration |
Verdict: the right ComfyUI model fits your hardware, workflow and licence, not the highest quality headline. Record the exact model, file, licence and review date.
Sources & Further Reading
Every source below was fetched on 2026-10-21.
| Claim | Source | Section or date |
|---|---|---|
| Supported families, node graph, quantized models, 4GB claim, GPL-3.0 | ComfyUI (GitHub) | README feature list; repo metadata |
| GGUF not natively supported | Models (ComfyUI docs) | Common issues; undated |
| FLUX.1 licences | black-forest-labs/flux (GitHub) | Open-weight models table; undated |
| FLUX.2 licences | black-forest-labs/flux2 (GitHub) | Model overview |
| FLUX 3 Image Klein and licence name | black-forest-labs/flux-3-image-klein (GitHub) | Model overview; License; repo pushed 2026-10-17 |
| Stable Diffusion 3.5 Community License | Community License Agreement (Stability AI) | Last updated July 5, 2024 |
| Ideogram 4 weights licence | Ideogram licensing | Licensing tiers; undated |
Left out after checking sources: any minimum VRAM requirement, because the official system-requirements page gives none, and the terms of the FLUX Kommunity License, which sits behind a login.
How we weigh evidence, and what terms like AGI and ASI mean on this site, is explained in our methodology.
Frequently asked questions
Which open-weight image model should I run in ComfyUI?
Choose the open-weight model by licence, hardware and workflow, not by headline quality. ComfyUI lists many supported families, but FLUX.1 schnell is Apache 2.0 while FLUX.1 dev is non-commercial, and GGUF files need a community node, so read the exact model licence first.
Can I use FLUX.1 dev commercially?
Its default licence is non-commercial. The same non-commercial licence covers FLUX.1 tools such as Fill, Canny, Depth and Kontext dev. Black Forest Labs points to a separate commercial licensing page, so review those current terms before using any of these weights in a commercial product.
What is GGUF quantization in ComfyUI?
GGUF is a file format for quantized models that can reduce memory use. ComfyUI documentation says it does not natively support GGUF and points to community custom nodes such as ComfyUI-GGUF. Use loaders and files that match each other, and expect possible changes in quality or speed.