FLUX.1 is the text-to-image model family that Black Forest Labs launched on August 1, 2024, as its debut product. Built by a team that had created VQGAN, Latent Diffusion, and the Stable Diffusion models, FLUX.1 scaled a hybrid architecture - combining transformer designs with flow-matching diffusion - to 12 billion parameters. On release it was widely judged to set a new bar for open image generation, particularly on prompt adherence, text rendering within images, and human anatomy, long-standing weak spots for earlier open models.
FLUX.1 shipped in three tiers that map directly onto a commercialization strategy. FLUX.1 [pro] is a closed, API-only model for top-end quality; FLUX.1 [dev] is an open-weight model offered for non-commercial use and distilled from pro; and FLUX.1 [schnell] is a fast, four-step model released under the permissive Apache 2.0 license for local and personal use. The lab later extended the line into image editing and video generation. Specific variants beyond the original three reflect Black Forest Labs’ announcements as of the verification date.
Why business readers should care: FLUX.1 reconcentrated the original Stable Diffusion talent and quickly became a default open base model for image generation, while its open-and-closed tiering became a common template for how generative-media startups release models.