fal and Black Forest Labs launch FLUX 3 Max, a video model they say renders a 10-second clip in about 12 seconds

German AI lab Black Forest Labs (BFL) and generative media platform fal introduced the FLUX 3 Max video model on October 19. It is built on FLUX 3, BFL’s video model with native audio released on August 4. According to fal’s launch page, fal Research added post-training, inference optimization and systems engineering, and FLUX 3 Max is served through fal. It launched with two endpoints, text-to-video and image-to-video. VentureBeat reports clips of 5 to 20 seconds at 720p, with optional upscaling to 1080p and audio that can be turned on or off.

By fal’s own measurement, a 10-second 720p clip takes 12.45 seconds on average. The same clip takes 277.80 seconds on Seedance 2.5 and 126.44 seconds on the base FLUX 3, which works out to roughly 22 times faster than Seedance 2.5 and 10 times faster than FLUX 3. fal says the model ties Seedance 2.5 for first place on Artificial Analysis’s preliminary image-to-video arena at 1,139 Elo. The leaderboard VentureBeat checked on October 19 showed MiniMax H3 at 1,140 and both FLUX 3 Max and Seedance 2.5 at 1,138. The confidence intervals overlap, so a clear quality lead is hard to establish.

List price is $0.17 per second of 720p video, or $1.70 for a 10-second clip. Until November 7 a 75% discount brings it to $0.0425 per second, and $0.0725 for output upscaled to 1080p. fal notes that speeds vary with settings and operating conditions. BFL says open-weight access to a FLUX 3 Dev model is coming soon.