Can you sell images from open-weights models? Licences compared
Open weights do not mean commercial use. Ideogram 4, Qwen-Image, FLUX.2 dev and LTX-2.5 differ on selling outputs. What each page says, and hosted rows.

It depends on the model, and "open weights" settles nothing. Among the models in this week's news, Qwen-Image's weights are Apache 2.0, while Ideogram 4's public weights are non-commercial, with a paid tier that adds commercial use. This is a reading of vendor pages, not legal advice.
Every row below is from the vendor's own page on 2026-10-03. If a licence matters to a contract, read the full text.
What does each licence say?
The summary lines are deliberately short; the cards hold conditions I do not repeat.
| Model | Licence as shown | Commercial use of outputs |
|---|---|---|
| Ideogram 4 | Non-commercial free tier; self-serve commercial tier at $300 a month | Only on the paid commercial tier |
| Qwen-Image | Apache 2.0 | Permitted under Apache 2.0 terms |
| FLUX.2 dev | Open weights, non-commercial (BFL pricing page) | Needs a separate licence from BFL |
| LTX-2.5 | LTX community licence, gated, under $10M revenue | Under the revenue threshold |
| MiniMax H3 | Community licence; application form for USA, EU, UK, South Korea | Per the licence and the application |
Why does Apache on the code not cover the weights?
A repository can carry one licence while the files it downloads carry another. Ideogram's licensing page describes Apache 2.0 for the inference code on GitHub, while the weights are under the Ideogram 4 Non-Commercial licence on the Hugging Face card. Reading only the repo licence is the common mistake; check the model card as well.
Does a revenue threshold matter to you?
Two licences here key on company size. LTX-2.5's community licence is shown for under $10M revenue, so a larger company needs another arrangement. Your own tax filing, not this post, decides which side of the line you are on.
How is a hosted call different?
A hosted call is governed by the provider's terms, not by the weights licence. That is why a team may test a family on Sume's hosted rows before choosing to download. Sume does not tell you which build serves an id, and its Image API has no custom-weights field, so you cannot bring a licensed checkpoint to it.
- Check the model card, not just the repository licence.
- Save the licence text and the date you read it with the project.
- Look for revenue thresholds and application forms.
- Treat gated downloads as agreements you accepted personally.
- Ask counsel when an ad or a client contract depends on it.
Sources
Related posts
- Is Ideogram 4.0 open source? Apache code, non-commercial weights
- Qwen Image license: is Qwen-Image open source under Apache 2.0?
- LTX-2 license: the $10 million annual revenue threshold
- FLUX.2 klein 4B Apache 2.0 vs 9B non-commercial, and Sume ids
- MiniMax H3 license: written authorization above 20M USD revenue
More in Models
- Cartesia voices speak up to 25 languages: voice plus language on Sume
Cartesia says 50+ library voices speak up to 25 languages natively. On Sume you send a voice id and a language code per job, and you test each pair.
- Cartesia Sonic 3.6: 44 languages or 61 locales, which to quote
Cartesia states 44 languages on its Sonic page and 61 locales in the Sonic 3.6 launch post. How to quote it, and what the Sume language field accepts.
- ChatGPT Try On from a screenshot: the same edit through an API
ChatGPT Try On starts from a selfie plus a product screenshot. Do the same edit with openai/gpt-image-2.5 on Sume: two references, one prompt, one Python call.
- Chinese, Hindi and Russian text to speech API: Sume zh, hi and ru
Sume's Voices library has zh, hi and ru tags. How to request each, what Eleven v4 lists, and the one field that stops an English-sounding read.
Written by Sume