Black Forest Labs Unveils FLUX 3: Multimodal Model Spanning Video, Image, Audio, and Robotics

FLUX 3 outperforms Grok, Kling, Runway, and Luma in early video benchmarks. The same backbone powers FLUX-mimic, a robotics model already deployed at Audi factories.

Black Forest Labs has released FLUX 3, a multimodal foundation model that jointly learns from images, video, and audio within a single unified architecture. The model is now available in Early Access.

FLUX 3 can generate videos up to 20 seconds long with native audio, supporting text-to-video, image-to-video, video-to-video, keyframe-to-video, and multilingual dialogue. Agentic chaining allows assembling individual clips into multi-shot sequences lasting several minutes.

In preliminary evaluations, FLUX 3 was preferred over Grok Imagine Video in 69% of comparisons, Kling v3 Pro in 60%, and Runway Gen-4.5 in 77%. It beat Luma Ray 3.2 in 93% of matchups. The model showed particular strength in capturing human facial expressions, synchronizing audio with physical events, and producing multilingual output.

The same model backbone also powers FLUX-mimic, a video-action model built with mimic robotics and deployed at Audi. FLUX-mimic handles soft-body manipulation tasks -- seals, cables, and flexible materials -- that conventional automation has never been able to touch, running at 101ms reaction time on a single NVIDIA RTX 5090.

FLUX 3 is built on BFL's Self-Flow approach for aligning multimodal generation and understanding. The company plans to roll out FLUX 3 Video, FLUX 3 Image, and an open-weight FLUX 3 Dev over the coming weeks.

Sources: Black Forest Labs -- Hacker News Discussion