FLUX.2 [dev]
Black Forest Labs · European Union · FLUX.1 [dev] Non-Commercial License
Commercial use: No — not licensed for commercial use
NO. Non-commercial licence covering the weights and the images produced. Commercial output needs a paid Black Forest Labs licence. Freely downloadable and freely used in breach — this remains the most common licensing mistake in AI image generation.
Note that FLUX.1 [dev] Non-Commercial License is not an OSI-approved open-source licence. The weights are downloadable, which is not the same thing.
Excellent image generation, freely downloadable, and the images it produces are not licensed for commercial use.
Why it should matter to you
The most expensive licensing mistake in AI image generation, because the output is the deliverable. Marketing assets, client work, product imagery, a website hero — all commercial use, all requiring a paid Black Forest Labs licence. Teams discover this after the campaign has run.
How it connects to our work
Where commercial output is required and budget is not, FLUX.1 [schnell] is Apache 2.0 from the same lab and clears the same bar for most work. We make that swap routinely on design engagements.
Advantages
- The best open-weight image quality currently available
- Large LoRA and fine-tune ecosystem
- Excellent prompt adherence and text rendering
Disadvantages
- Outputs are not commercially usable without a paid licence — this includes client work, marketing and anything on a company website
- Large: 32B needs real hardware
- Slower than the klein variants
Reach for it when
Research, personal work, and evaluating whether the paid licence is worth buying.
Where it falls down
Any commercial deliverable, as a matter of licence rather than capability. Use FLUX.2 klein 4B under Apache 2.0, or buy the BFL licence.
Running it
24 GB GPU at Q4. Check the licence before you ship anything it made. See the hardware sizing tables for how that maps to specific chips and cards, and the quantisation guide for what you give up at each bit width.
Free tiers carrying this model
| Provider | Type | The catch | Live limits |
|---|---|---|---|
| OpenRouter | Aggregator | Free-pool membership changes without notice — a model you built on can stop being free, and the list above will drift. Data passes through OpenRouter and then the upstream provider, so check the privacy and routing settings before sending anything sensitive. | check → |
Jurisdiction: European Union
Mistral is the only EU-headquartered lab shipping frontier-adjacent open weights. Contracting with an EU entity removes the international-transfer question entirely rather than solving it with a framework. Most of the Mistral open range is Apache 2.0, so you can also self-host and remove the vendor from the data path altogether.
Watch for: EU jurisdiction is not the same as EU data residency, and neither is the same as EU-only sub-processors. Ask for all three in writing if it matters to you. Also note that Codestral ships under a non-production licence — being a European vendor does not make every model commercially free.
Frequently asked
Can I use FLUX.2 [dev] commercially?
NO. Non-commercial licence covering the weights and the images produced. Commercial output needs a paid Black Forest Labs licence. Freely downloadable and freely used in breach — this remains the most common licensing mistake in AI image generation.
What hardware do I need to run FLUX.2 [dev]?
24 GB GPU at Q4. Check the licence before you ship anything it made. Weights alone are roughly 18 GB at Q4_K_M and 34 GB at Q8_0. Add KV cache on top of that, which grows with your context length.
What licence is FLUX.2 [dev] released under?
FLUX.1 [dev] Non-Commercial License. Outputs and weights are for non-commercial use. A paid Black Forest Labs licence is required for commercial output. This is not an OSI-approved open-source licence, whatever the marketing says.
Where can I use FLUX.2 [dev] for free?
Free tiers carrying it include OpenRouter. Limits differ per provider and change often, so check each provider's own limits page. You can also self-host the weights, which has no rate limit at all.
Similar models
Mistral Large 3 675B
Multi-GPU datacentre, or hosted on La Plateforme.
Shieldstral 1.0 3B
4 GB. Runs on almost anything.
DeepSeek-V4-Flash
Hosted, or 256 GB+ unified memory at Q4. Not a laptop model.
GLM-5.3-Flash
80-96 GB at Q4_K_M. A workstation or Ultra-class Mac.
Wiring FLUX.2 [dev] into something real?
We build the evaluation harness, the failover and the cost ceilings around a model like this, so it survives contact with production.
Maps to AI graphic design and AI automation development. Or see it working: our case studies.