fal launches FLUX 3 Max, a video model 22x faster than Seedance 2.5
fal has launched FLUX 3 Max, a video generation model post-trained by fal Research in collaboration with Black Forest Labs. According to fal, it reaches top-tier image-to-video quality and creates a 10-second video in 10.95 seconds.
- fal launched FLUX 3 Max, post-trained with Black Forest Labs from the FLUX 3 foundation model.
- It scored 1,139 Elo across 4,868 votes on Artificial Analysis Video Arena in preliminary image-to-video evaluations.
- fal says it generates a 10-second video in 10.95 seconds, over 22 times faster than Seedance 2.5 at the same reported Elo.
- Compared with the original FLUX 3, Elo rose from 1,089 to 1,139 and latency fell from 126.44 to 10.95 seconds.
- Available from October 19, 2026 on fal's platform, at 75% off for the first three weeks.
Quality and speed figures
FLUX 3 Max builds on the FLUX 3 foundation model through specialized post-training, reinforcement learning and inference optimization. In preliminary independent human preference evaluations on Artificial Analysis Video Arena, the model reached a 1,139 Elo rating across 4,868 votes. According to fal, that places it among the highest-quality image-to-video models evaluated, with overlapping confidence intervals among the top-ranked models.
The company's own latency benchmarks show the model generating a 10-second video in 10.95 seconds. Seedance 2.5, which fal says has the same reported Elo rating, needed 277.80 seconds, which makes FLUX 3 Max over 22 times faster.
Against the original FLUX 3, fal reports two gains:
- Image-to-video Elo rose by 50 points, from 1,089 to 1,139.
- Generation latency fell from 126.44 seconds to 10.95 seconds, which is 11.5 times faster.
How the two companies split the work
fal describes the project as a collaboration between two research teams. Black Forest Labs contributed foundation model development, while fal Research handled post-training, reinforcement learning, inference optimization and systems engineering. According to fal, the team worked directly with FLUX 3 to improve visual fidelity, motion coherence, prompt adherence and overall generation performance.
The model also draws on earlier work. fal says FLUX 3 Max builds on the research behind H3 Max, its previously released video model, which produced video in 3.49 seconds in fal's benchmark while achieving top-tier image-to-video quality in independent evaluations. FLUX 3 Max applies the same approach to a different foundation model.
Robin Rombach, co-founder and CEO of Black Forest Labs, said: "We built FLUX 3 as a foundation for others to build on." He added that fal "took our model and customized it for the developers and enterprises on their platform."
Kaan Oktay, Head of Research at fal, said: "Model performance doesn't stop improving when base model training ends." Burkay Gur, CEO of fal, said the company believes some of the most important advances in generative AI will come from optimizing models and the systems that run them together.
Availability and what it means for developers
FLUX 3 Max is available from October 19, 2026, through fal's inference platform. Developers get text-to-video, image-to-video and keyframes-to-video generation through a unified API. To mark the launch, fal is offering the model at 75% off for the first three weeks.
The company calls this the first iteration of FLUX 3 Max and says fal Research is continuing to post-train it, with further gains in visual quality, generation speed and overall performance underway.
For developers building video features, the pitch is a reduction in waiting time without a drop in rated quality. All speed figures come from fal's own benchmarks, and the Elo rating is described as preliminary, so independent results may shift as more votes are collected.
