On July 22, 2026, Black Forest Labs (BFL) released FLUX 3 — the successor to FLUX.2 and the first major model to unify image and video generation in a single architecture.
What’s new
- Unified image + video. One model generates both native 4K stills and 10-second 4K video clips. No separate image and video endpoints.
- Real-time interactive editing. Users can adjust compositions, styles and elements interactively during generation — a workflow closer to creative tools than batch-prompt APIs.
- 10-second 4K clips. The video output matches dedicated video models on duration and resolution, a notable jump from BFL’s image-only lineage.
Access and pricing
FLUX 3 launched with a beta API — no open weights at launch. BFL has signaled that open-weight releases may follow (continuing the FLUX.2 [dev] tradition), but nothing is confirmed.
Pricing details were not finalized at launch. FLUX.2’s hosted routes ran from ~$0.014/image; expect FLUX 3’s video generation to carry a per-second premium.
Why it matters
- The image-video wall is falling. FLUX 3 is the clearest signal yet that the next generation of media models won’t separate still images from motion — they’ll be the same model with different output modes.
- Real-time editing is the differentiator. Most video models are batch processes: prompt, wait, get a clip. Interactive editing during generation puts FLUX 3 closer to creative-tool territory.
- Open weights are the open question. FLUX.2 [dev] became the de facto open standard for image generation pipelines. If FLUX 3 weights ship under similar terms, it could become the open standard for video too — which is why we’ve flagged it in both our image and video rankings.
Full specs: FLUX 3 model page. See also: Best AI Image Generators and Best AI Video Generators.