FLUX.2 dev commercial use: what the model card says
FLUX.2 [dev] weights carry the FLUX non-commercial license, yet the card says outputs can be used commercially as that license describes. The exact wording.

The FLUX.2 [dev] weights are under the flux-non-commercial-license, but the model card also says generated outputs "can be used for personal, scientific, and commercial purposes, as described in the FLUX [dev] Non-Commercial License". So the answer is in the license text, not the card summary. Read that license before you ship anything.
This post quotes the FLUX.2-dev model card on Hugging Face, read 2026-09-29. It is not legal advice. The Sume side is from Image generation.
What does the model card say?
| Point | What the card says |
|---|---|
| Model | A 32 billion parameter rectified flow transformer for generating, editing and combining images from text instructions |
| License label | flux-non-commercial-license |
| Outputs | Can be used for personal, scientific, and commercial purposes, as described in the FLUX [dev] Non-Commercial License |
| Access | Publicly accessible, but you must accept the conditions (the FLUX Non-Commercial License Agreement and the Acceptable Use Policy) to access files |
Can I run FLUX.2 dev on a consumer GPU?
The card points to the diffusers docs for local deployment on a consumer graphics card such as an RTX 4090 or an RTX 5090. Its example loads a 4-bit quantized model with a remote text encoder on an RTX 4090. It gives no single minimum VRAM number, so this post gives none.
What is FLUX.2 [dev] built for?
The card lists its key features: text-to-image, single-reference editing and multi-reference editing, character, object and style reference without additional training in one model, and training with guidance distillation. It says the open weights are meant to drive scientific research and empower artists to develop workflows, and that the model is available in ComfyUI and Diffusers.
Those are the [dev] model's features. Do not read them as a list of what a hosted FLUX.2 endpoint accepts; check the hosted model's own page.
Is FLUX.2 available through a hosted API?
Sume's image docs use black-forest-labs/flux.2-pro as a model id in a request, and the same page says GET /v1/images/models lists the available models. That is FLUX.2 [pro], a hosted model called with an API key; the docs do not say it uses the downloadable [dev] weights, so do not assume the [dev] license terms apply to a hosted call, and read the terms of the service you use. FLUX.2 API pricing covers what a hosted Pro or Flex call costs.
curl "https://api.sume.com/v1/images/models" \
-H "Authorization: Bearer $SUME_API_KEY"What should I check before commercial use?
- The full license text linked from the card, since the card only summarizes it.
- The Acceptable Use Policy the card tells you to acknowledge.
- The terms of any hosted service you use instead of the weights, which are separate from the card.
Sources
Related posts
More in Models
- FLUX.2 flex text rendering: prompt tips and the Sume model id
Black Forest Labs positions FLUX.2 [flex] for text and fine detail. How its typography guidance reads, and how to call black-forest-labs/flux.2-flex on Sume.
- FLUX.2 hex color prompt: how to write it, with an API call
Black Forest Labs says FLUX.2 matches hex codes in prompts. The syntax it documents, the limits it admits, and a flux.2-pro call on Sume.
- FLUX.2 pro vs Nano Banana Pro API: request specs side by side
Request parameters that differ between black-forest-labs/flux.2-pro and google/nano-banana-pro on Sume: resolution tiers, aspect ratios, references and price.
- FLUX.2 prompt guide: structure, length and no negatives
Black Forest Labs' FLUX.2 prompt rules: subject, action, style, context; 30-80 words; no negative prompts. Which apply to flux.2-pro on Sume.
Written by Sume