fal releases H3 Max, a post-trained MiniMax H3 video model claiming 3-second generation for 5-second clips
- fal released H3 Max, a post-trained video model built on open-weights MiniMax H3, for text-to-video and image-to-video generation via its API, Playground, and Agent.
- fal says H3 Max generates a five-second video in about three seconds of wall time, claiming roughly 35 times the throughput of MiniMax's official H3 endpoint and 15 times the average speed of 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; Artificial Analysis ranks fal's H3 first for image-to-video with audio at 1,201 Elo across 2,177 samples.
- fal post-trained MiniMax H3 on new data with its reinforcement-learning framework, using human preference studies for visual quality, prompt understanding, and aesthetics while co-designing inference optimizations with the evolving model.
- During its first launch week, fal prices 768p H3 Max output at $0.04 per second, a 50% promotional discount.