Black Forest Labs

Founding & background

Black Forest Labs (BFL) was founded in 2024 by former Stability AI researchers — it's based in Freiburg im Breisgau, Germany, and was founded by former employees of Stability AI. The founding team pioneered latent diffusion, the technology behind Stable Diffusion, and BFL now operates as a roughly 100-person team building the foundation layer for visual intelligence, working out of both Freiburg, Germany, and San Francisco, USA. Founder Robin Rombach was lead author on "High-Resolution Image Synthesis with Latent Diffusion Models" (CVPR 2022 Oral) — the paper that established latent diffusion as the efficient architecture behind Stable Diffusion. The team is noted for having contributed to models cited over 200,000 times academically.

Funding & valuation

BFL is valued at $3.25 billion and has raised more than $450 million from investors including a16z, AMP, Salesforce Ventures, NVIDIA, General Catalyst, Adobe Ventures, Figma Ventures, Canva, and Deutsche Telekom's T.Capital.

Product history — the FLUX line

• FLUX.1 launched in August 2024 as a text-to-image model, with licensing split across variants: Apache License for Schnell and Klein, non-commercial for Dev, and proprietary for Pro and Flex.
• FLUX.2 became the stable release as of November 25, 2025. A key feature of FLUX.2 is multi-reference support, letting it process up to 10 images at once while maintaining character and style consistency.
• BFL's FLUX.1 models became widely used as a base for community fine-tuning within the open diffusion ecosystem.

FLUX 3 — the big July 2026 launch

On July 23, 2026, BFL launched FLUX 3, a single set of weights trained jointly across images, video, and audio, with the same architecture extended to predict robot actions — making it the first major generative model where 20-second video with native synchronized audio, advanced image editing, and factory-floor robotic control all run through one shared backbone.

• Architecture: Audio and video are generated together in a single pass for better causal alignment — sound timing matches impact frames, dialogue matches lip movement, and ambient noise matches the scene. The underlying method, called Self-Flow, was published by BFL in March 2026 to unify generation quality and representation quality within one framework.
• Product lines: FLUX 3 ships across four product lines: FLUX 3 Video, FLUX 3 Image, FLUX 3 Action, and FLUX 3 Dev.
• Rollout status: FLUX 3 Video and FLUX 3 Action (also called FLUX-mimic) are in early access now via API and private weights for select partners; FLUX 3 Image is rolling out "in the coming weeks" with improved complex-prompt handling and multilingual text rendering; and FLUX 3 Dev, the open-weight tier, is planned for later in 2026 to enable local deployment and customization. No pricing has been disclosed for any tier yet.
• Robotics angle: The robotics product is already running production tests inside Audi's factories.
• Benchmark performance: In early preference testing on 10-second, 720p text-to-video clips with audio, BFL reports FLUX 3 was preferred over Luma Ray 3.2 in 93% of comparisons, Runway Gen-4.5 in 77%, Grok Imagine Video in 69%, Kling v3 Pro in 60%, and roughly a coin flip (52%) against both Seedance 2.0 and Google's Gemini Omni Flash. (Worth noting: these are vendor-reported, preliminary results from BFL's own testing.)

Market position

This launch puts the $3.25 billion startup in direct competition with OpenAI's physical-AI ambitions, and marks BFL's expansion from a pure image-generation company into video, audio, and robotics — a notably broad bet for a company that, until this release, had shipped image-only models.

Want me to go deeper into FLUX 3's technical architecture (Self-Flow), how it stacks up against specific competitors, or the FLUX.1/FLUX.2 image model details?