FLUX 3 Max brings a latency-focused variant to fal’s video API
fal’s model page presents FLUX 3 Max as a post-trained variant of FLUX 3 Video, developed in collaboration with Black Forest Labs. The text-to-video endpoint accepts a written prompt without requiring a starting image and supports clips from 5 to 20 seconds, 720p or 1080p output, multiple aspect ratios, optional audio, and selectable reasoning effort.
fal reports that a 10-second video takes approximately 11 seconds in its benchmark configuration, more than 25 times faster than Seedance 2.5 in the tested comparison. The page also reports a joint first-place result in a preliminary Artificial Analysis image-to-video evaluation. These claims describe different evaluations: the ranking is not a text-to-video result, and actual generation time varies with settings and operating conditions.
The release’s practical proposition is faster iteration on prompted scenes, not a general guarantee of generation speed or quality. The page lists commercial use and exposes the model through an API, giving builders a route to test it in creative workflows. Source: https://fal.ai/models/blackforestlabs/flux-3-max/text-to-video