Black Forest Labs announces FLUX 3 as its first multimodal frontier model, with only the video variant in early access
FLUX 3 was announced on 23 July as Black Forest Labs' first multimodal frontier model, released on a phased rollout with the video variant alone in early access. An independent image lab moving to multimodal frontier framing is a statement about where the category floor now sits.
Black Forest Labs built its position on image generation at a time when that was a category. The move to describe FLUX 3 as a multimodal frontier model concedes that image-only is no longer a defensible product boundary, because every general model now does images adequately and the differentiated work has moved to video and cross-modal editing.
A phased rollout that leads with video and holds the rest back is the opposite of the usual sequencing, and it reads as capacity-driven. Video inference is dramatically more expensive per output than image inference; releasing it first to a controlled cohort is how a lab without hyperscaler backing manages that cost while still planting a flag.
The comparison that matters is against Gemini Omni, announced into the same month with YouTube-scale distribution attached. An independent lab and a hyperscaler shipping comparable capability in the same window is exactly the setup where distribution decides the outcome regardless of which model is better.
Digital Applied — Seven Days, Seven Model Releases → · AI Avatar Tech — The Ultimate Guide to AI Video Models in 2026 →