fal has debuted FLUX 3 Max, a video generation model post-trained by fal Research in collaboration with Black Forest Labs on the FLUX 3 foundation model. The release, announced October 19, 2026, pairs Black Forest Labs’ foundation model with fal’s research in post-training, reinforcement learning, and inference optimization.
The model targets a sweet spot the video AI market has been chasing: top-tier image-to-video quality at near real-time speed. fal reports a 1,139 Elo rating in preliminary human preference evaluations on Artificial Analysis’s Video Arena, placing it among the highest-quality image-to-video systems tested. Its benchmark measured a 10-second video generating in 10.95 seconds — which the company describes as 22 times faster than Seedance 2.5 at the same reported quality rating.
FLUX 3 Max is served through fal’s inference platform with text-to-video and image-to-video endpoints at 720p. Launch pricing is $0.0425 per second of output video, a 75% discount off the $0.17 list price running through November 7, 2026, with per-second billing and no subscription.
fal Head of Research Kaan Oktay said the team achieved a 50-point Elo improvement over the original FLUX 3 while making inference more than 11 times faster, and that further iterations are already underway.
Explore video AI tools in our Flux, Runway, and Midjourney profiles, or compare leading video tools in our Runway vs Pika guide.
Source: Unite.AI
