FLUX.3 [dev] is coming with open weights for image, video, audio, and robot action prediction

multimodalart · x · 2026-07-24

The post says FLUX.3 [dev] is coming with open weights and support for image, video, audio, and robotic action prediction.

It also points to the model’s self-flow architecture, described as a new evolution of flow matching. The author praises the team for pushing architectural innovation in addition to scale.

Related event: Black Forest Labs Launches FLUX 3: A Unified Multimodal Foundation Model(26 posts)→

Original post →

More from Multimodal

Multimodal channel →