fal launches H3 Max, a post-trained MiniMax H3 video model optimized for sub-3-second 5-shot generation

fal Research post-trained MiniMax’s open-weights H3 video foundation with new data and inference optimization, targeting throughput rather than just quality. fal claims a 5-second video in under 3 seconds (~35x faster than the official H3 endpoint), #1 across overall quality, prompt understanding, and aesthetics in fal’s head-to-head human-preference Elo over 12 models.