FLUX & Black Forest Labs Statistics: Downloads, Valuation & Adoption
FLUX (Black Forest Labs): 400M+ model downloads, $3.25B valuation after a $300M Series B (Dec 2025), founded 2024 by original Stable Diffusion researchers, embedded in Adobe, Canva, Meta, Microsoft products.
FLUX, built by Freiburg-based Black Forest Labs, became the fastest-rising model family in AI image generation. Founded in 2024 by the original creators of Stable Diffusion, the company crossed 400 million model downloads and a $3.25 billion valuation within two years - while keeping many model weights open. FLUX models now power image features inside Adobe Firefly, Canva, Meta, and Microsoft products.
Key statistics
FLUX models have been downloaded more than 400 million times, ranking among the most widely deployed image generators in the world.
Black Forest Labs was valued at $3.25 billion in December 2025 after a $300 million Series B round.
The FLUX.1 family uses a 12-billion-parameter rectified flow transformer architecture, larger than SDXL’s ~3.5B parameters.
FLUX sits inside products from Adobe (Firefly partner model), Canva, Meta, Microsoft, ElevenLabs, and Vercel.
FLUX.2 delivers resolutions up to 4 megapixels with improved in-image text rendering and multi-reference consistency.
Which FLUX variant fits which use case?
The FLUX family splits along a license line: schnell is Apache 2.0 (free for any use), dev is open-weight but non-commercial without a license, and the pro tier is API-only. That split is what let FLUX capture both the self-hosted Stable Diffusion crowd and enterprise API buyers at once.
| Variant | License / access | Best for | Main tradeoff |
|---|---|---|---|
| FLUX.1 schnell | Apache 2.0 open weights | High-volume self-hosted pipelines | Lowest quality of the family; 4-step distilled model |
| FLUX.1 dev | Open weights, non-commercial | Local experimentation, LoRA training | Commercial use requires a paid license |
| FLUX.1 / 1.1 pro | API only | Production quality via API | No self-hosting; per-image cost |
| FLUX.2 | API + select open variants | 4MP output, multi-reference consistency, text rendering | Newest tier; higher compute cost |
Variant and license details from Black Forest Labs official documentation (blackforestlabs.ai), June 2026. Verify license terms before commercial self-hosted deployment.
Why did FLUX adoption grow so fast?
FLUX combined open-weight distribution (the growth engine that made Stable Diffusion ubiquitous) with frontier-level quality that benchmarks ahead of most closed models.
Black Forest Labs was founded in 2024 by former Stability AI researchers Robin Rombach, Andreas Blattmann, and Patrick Esser - the team behind the original Stable Diffusion paper.
The December 2025 Series B brought in $300 million, predominantly from US investors - Europe invests roughly 12x less in late-stage AI rounds.
In a 2025 mixed-methods study, participants correctly identified FLUX.1-dev images as AI-generated in only 29% of cases - below chance, and the lowest detection rate of all models tested.
Unlike Midjourney or GPT Image, many FLUX variants (schnell, dev) ship as open weights - making FLUX the de facto successor to Stable Diffusion for self-hosted and API-aggregated deployments.
How we compiled this data
Figures combine Black Forest Labs’ official disclosures, funding reporting around the December 2025 Series B, Hugging Face download counters, and the Frontiers in Artificial Intelligence detection study (2025), which we cite directly rather than via press coverage. Download totals were sanity-checked against per-variant Hugging Face counts. Last full review: June 12, 2026.
Before you cite these numbers
- The 400M+ download figure aggregates all variants and re-downloads across mirrors; it is not 400 million distinct users. API-only pro usage is not public at all.
- The $3.25B valuation is a private-market price from one funding round, not a traded value.
- The 29% detection rate comes from one 2025 study with a specific participant pool and prompt set; detection rates vary widely across study designs.
- FLUX benchmark superiority over SDXL refers to out-of-the-box base models. A fine-tuned SDXL or SD 3.5 pipeline can outperform base FLUX for narrow domains.
Frequently asked questions
How many times has FLUX been downloaded?
FLUX models have been downloaded more than 400 million times across Hugging Face and other distribution channels, making the family one of the most widely deployed image generation models in the world.
What is Black Forest Labs worth?
Black Forest Labs was valued at $3.25 billion in December 2025 following a $300 million Series B round. The company was founded in 2024 by the original creators of Stable Diffusion.
Is FLUX better than Stable Diffusion?
Independent benchmarks consistently rank FLUX above SDXL in prompt adherence, detail quality, and text rendering out of the box. In one 2025 detection study, FLUX.1-dev images fooled human evaluators more often than any other model - only 29% were correctly identified as AI.
Which products use FLUX?
FLUX models are integrated into Adobe Firefly (as a partner model), Canva, Meta and Microsoft products, ElevenLabs, and Vercel, and are available through APIs like fal.ai - including the FLUX models available on Morphed.
Is FLUX better than DALL·E?
For photorealism and detection-resistance, benchmarks favor FLUX: its images were correctly identified as AI only 29% of the time in a 2025 study, the lowest rate measured. DALL·E’s successor (GPT-4o native generation) wins on instruction-following and conversational editing. The practical answer depends on whether you need API control and open weights (FLUX) or ChatGPT integration (OpenAI).
Sources
Figures on this page are compiled from the following publishers and reports. Where sources disagree, we present the range and note the methodology difference.