fal releases H3 Max, a post-trained MiniMax H3 video model claiming #1 benchmark rankings
- fal released H3 Max, a post-trained model built on open-weights MiniMax H3, for text-to-video and image-to-video generation through its API, Playground, and fal Agent.
- fal says H3 Max generates a five-second video in about three seconds, roughly 35 times the throughput of the official MiniMax H3 endpoint and an average of 15 times faster than comparable-quality models in its tests.
- Design Arena ranks H3 Max first on its recent image-to-video leaderboard at 1,341 Elo, ahead of official MiniMax H3 at 1,333, according to the announcement.
- Artificial Analysis ranks fal's H3 model first on its Image-to-Video with Audio leaderboard, with 1,201 Elo across 2,177 samples, the company says.
- fal post-trained H3 with new data and an in-house reinforcement learning framework, while co-designing inference optimizations that it retained only when human preference tests preserved quality gains.