Black Forest Labs (BFL) is a German artificial intelligence company based in Freiburg im Breisgau, Germany, specializing in generative AI for image and video creation. Founded in 2024 by former Stability AI employees, the company develops the Flux family of text-to-image and image-to-image models, which generate images from natural language prompts and support editing capabilities. BFL positions itself as a research lab while offering both open-source and proprietary models through APIs and partnerships with major technology firms.
The company emerged from academic research at LMU Munich, where its founders contributed to the development of Stable Diffusion. Black Forest Labs has attracted significant venture capital and strategic investments, and its models have been integrated into platforms such as Grok, Le Chat, and Adobe Photoshop. By 2026, the company expanded into video generation and robotics, reflecting its ambition to advance multimodal AI.
Founding and Funding
Black Forest Labs was founded in 2024 by Robin Rombach, Andreas Blattmann, and Patrick Esser, along with seven other co-founders. All three principal founders had previously worked as research assistants at LMU Munich under Björn Ommer, where they researched AI image generation and published results in April 2022 that led to the creation of Stable Diffusion. They later joined Stability AI before leaving to establish BFL.
The company received an initial investment of US$31 million from investors including venture capital firm Andreessen Horowitz, Brendan Iribe, Michael Ovitz, Garry Tan, and Vladlen Koltun. By the end of 2025, investor interest had surged, with AMP, Salesforce Ventures, Samsung, Adobe, Deutsche Telekom, and Canva offering US$450 million. BFL subsequently incorporated as a Delaware corporation, with investors also becoming customers who could use the AI for product images, advertising, brand design, e-commerce, and integration into tools like Adobe Photoshop.
As of a Forbes article in June 2026, the company had 70 employees and was still considered a startup, competing against better-funded Silicon Valley companies. In August 2026, DLF reported that BFL had 100 employees, with thirty already based in San Francisco and another thirty soon to join. Despite its growth, the company emphasized its identity as a research lab. In interviews, the founders declined to disclose their training data sources, data centers, or sustainability goals, and their transparency hub did not reveal data origins.
Development and Partnerships
In August 2024, Flux was integrated into Elon Musk's Grok chatbot developed by xAI, available as a premium feature on X (formerly Twitter). Four months later, in December 2024, Grok switched to its own text-to-image model, Aurora. On 18 November 2024, Mistral AI announced that its Le Chat chatbot had integrated Flux Pro as its image generation model.
On 21 November 2024, BFL announced the release of Flux.1 Tools, a suite of editing tools designed to work with existing Flux models. The tools included Flux.1 Fill for inpainting and outpainting, Flux.1 Depth for control based on extracted depth maps, Flux.1 Canny for control based on extracted canny edges, and Flux.1 Redux for mixing input images and prompts. Each tool was available in both Pro and Dev models.
In January 2025, BFL announced a partnership with Nvidia to include Flux models as foundation models for Nvidia's Blackwell microarchitecture. The company also released the Flux Pro Finetuning API for customization and fine-tuning, and partnered with German media company Hubert Burda Media for content creation.
On 29 May 2025, BFL announced Flux.1 Kontext, a suite of models enabling in-context image generation and editing, allowing users to prompt with both text and images. Alongside this, BFL Playground, an interface for testing Flux models, was released. On 31 July 2025, BFL announced Flux.1 Krea Dev, a model developed with Krea AI for better performance, varied aesthetics, and improved realism.
In September 2025, Adobe announced that Photoshop (beta) users could use Flux.1 Kontext Pro as a model for its generative fill tool. BFL also collaborated with Meta on Vibes, a video-generation app. On 25 November 2025, BFL released the Flux.2 model series, including Pro, Flex, Dev, and Apache 2.0-licensed Klein models, along with a Flux.2 variational autoencoder released as open-source software under Apache 2.0. This series claimed improvements in image reference, photorealism, typography, and prompt understanding.
Model Architecture and Licenses
Flux is a series of text-to-image models based on rectified flow transformer blocks scaled to 12 billion parameters. The models use a Transformer (architecture) architecture, which differs from earlier approaches like U-Net-based diffusion models. Flux.1 models were released under different licenses: Schnell (German for 'fast') as open-source software under Apache License, Dev as source-available software under a non-commercial license (with commercial licenses available from BFL), and Pro as proprietary software available only via API. Users retained ownership of generated outputs regardless of model used.
An improved flagship model, Flux 1.1 Pro, was released on 2 October 2024. Two additional modes were added on 6 November: Ultra, which generates images at four times higher resolution up to 4 megapixels without affecting speed, and Raw, which generates hyper-realistic images in the style of candid photography.
Flux.1 Kontext is a series with in-context image generation and editing capabilities, available in Max, Pro, and Dev models. Max is the highest quality model and can iteratively modify existing images using prompts, while Pro balances quality and speed. Dev is an open-weight model under a non-commercial license, similar to Flux.1 Dev.
Flux.2 models are based on latent flow matching architecture with Mistral AI's Mistral-3 model (24 billion parameters) for its vision-language model. As with Flux.1, Flux.2 models were released under different licenses: Klein as open-source under Apache License, Dev as source-available under non-commercial license, and Flex and Pro as proprietary APIs.
The models can be used online or locally via generative AI user interfaces such as ComfyUI, Recraft Studio, and Stable Diffusion WebUI Forge (a fork of Automatic1111 WebUI).
Video Generation and Robotics
In July 2026, BFL released FLUX.3, a text-to-video model, including FLUX3 x mimic, developed in collaboration with mimic robotics. FLUX3 x mimic is a video-action model running on robots, performing 'soft-body manipulation work' at Audi car production facilities. This marked BFL's expansion beyond still images into video and embodied AI applications.
Reception
According to a 2024 test performed by Ars Technica, the outputs generated by Flux.1 Dev and Flux.1 Pro were comparable with DALL-E 3 in terms of quality and fidelity. The models received attention for their open-source availability and high performance, positioning BFL as a significant player in the Generative AI landscape.
In June 2026, film director Martin Scorsese joined BFL as an adviser, which generated pushback from creative professionals and the Art Directors Guild, reflecting ongoing debates about AI's role in creative industries. The company's rapid growth and partnerships with major tech firms highlighted its influence, while its opacity regarding training data drew criticism from transparency advocates.
See Also
References
This article is based on information from public sources and company announcements as of 2026.