Flux 1.0 (Black Forest Labs): 12B parameters, flow matching, the new open source SOTA
In one sentence Black Forest Labs, founded by ex-Stability AI team, launches Flux 1.0 with flow matching architecture at 12 billion parameters, setting new open source standards on prompt adherence and visual quality.
Flux 1.0 arrived like an earthquake in the open source scene. The team behind it consists largely of ex-Stability AI researchers — the ones who built Stable Diffusion — who left and founded Black Forest Labs to do something even more ambitious.
The result is a model with 12 billion parameters (about 14 times SDXL) using a different architecture from all previous ones, called "flow matching." The text you give the model is followed much more precisely than previous models: if you say "three red apples on a wooden table," you get exactly that.
Released in three versions (one fully open, one for personal use, one commercial), Flux quickly replaced SDXL as the community's reference model.
Companies
Black Forest Labs
Tools
Flux 1.0, Flux.1-dev, Flux.1-schnell
Tags
Sources