FLUX 3: Black Forest Labs announces a multimodal flow model that surpasses Seedance 2.0 and Gemini

Black Forest Labs releases FLUX 3, its new generation of multimodal flow matching models. The lab claims superior results to Seedance 2.0 and Gemini on internal benchmarks. It remains to be confirmed by third-party evaluations.

The fact

Black Forest Labs releases FLUX 3, a new generation of multimodal models based on flow matching. The lab claims - on its internal benchmarks - superior results to Seedance 2.0 (ByteDance) and Gemini on directed image editing and long composition. FLUX 3 unifies text, image, and short video in the same backbone, with an open fine-tuning track for studios.

Our analysis

Since the controversy around SD3 in 2024, Black Forest Labs has operated as the European lab that Stable Diffusion should have become: open, technical, without product hype. FLUX 3 confirms that the multimodal race is no longer only happening at OpenAI, Google, and ByteDance. One point to validate: internal flow matching benchmarks do not yet reflect real usage in production pipelines (long shot coherence, style stability, typographic control). Wait for third-party evaluations (Artificial Analysis, Chatbot Arena vision) before concluding a shift.

To watch

The integrations of Adobe / Runway around FLUX 3, and the clarification of the license (open weights vs commercial restrictions). If the lab maintains openness, the gap with Seedance in China and Sora in the West narrows for independent studios and founders building on top of the image layer.