Black Forest Labs launches with the FLUX.1 image models

Background

Black Forest Labs was founded by Robin Rombach, Patrick Esser and Andreas Blattmann, who are described as the original creators of Stable Diffusion ^[2]. The company announced itself with this release of its first models ^[1]. A seed round is an early funding round, usually the first outside money a young company raises, and it is often used to build the first products. Black Forest Labs described its $31 million seed round as the funding that accompanied the launch ^[1].

What happened

On 1 August 2024 Black Forest Labs announced itself and released the FLUX.1 text-to-image models ^[1]. It raised a $31 million seed round led by Andreessen Horowitz ^[1].

The models came in three variants. FLUX.1 [pro] was offered through the API. FLUX.1 [dev] was open-weight, with a non-commercial license. FLUX.1 [schnell] was released under the Apache 2.0 license. All three variants had 12 billion parameters ^[1][2].

Key details

The models use a hybrid multimodal and parallel diffusion transformer, trained with flow matching ^[1]. A diffusion transformer combines the diffusion process, which generates images by removing noise step by step, with the transformer architecture.

Black Forest Labs claimed that FLUX.1 [pro] and [dev] beat Midjourney v6.0 and DALL·E 3 (HD) on quality and prompt following ^[1]. These are the company's own claims, made at launch.

The company said text-to-video was its next frontier ^[2]. This is a statement of plans at the time of launch, not a release.

Impact

The launch set up the company's first model family and its three-tier pattern of hosted, open-weight and commercial-friendly variants ^[1][2]. Later models, such as FLUX1.1 [pro] in October 2024, are recorded in the FLUX.1 entry in this archive. The launch date, the seed round and the three variant names are the points to check against the company's own announcement ^[1].