Crypto
Home›Crypto›Market Structure›Black Forest Labs launches FLUX 3 multimodal video mod…
Black Forest Labs launches FLUX 3 multimodal video model in early access
The model’s video and synced audio generation is already being used as the backbone for a robotics system Audi is testing on its production line.
Black Forest Labs has released FLUX 3 in early access, its first model designed to generate video rather than only still images. The system can produce clips up to 20 seconds long, generating synced audio alongside the visuals, including dialogue, sound effects, and ambient noise.
The company said FLUX 3 is built as a shared, multimodal backbone trained on images, video, and audio together in one system. In early evaluations, human reviewers preferred FLUX 3 over Runway Gen-4.5 in 77% of head-to-head comparisons, and over Luma Ray 3.2 in 93%.
Black Forest Labs also positioned FLUX 3 as more than a content tool, saying that predicting video helps the model learn underlying physics like weight, contact, and timing. The outlet Decrypt reports that, in the company’s view, that translates to better performance in physical tasks.
Beyond media creation, the same core video-prediction engine is used for FLUX-mimic, a robotics model built with mimic robotics and Zurich-based mimic robotics. According to Decrypt, Audi is already testing FLUX-mimic on production-line tasks such as fitting flexible door seals.
Looking ahead, only the open-weight FLUX 3 “Dev” version is planned for later in 2026, while Video and Action remain available via APIs and partner access for now. Decrypt added that Image is expected in the coming weeks.