fal's post-trained MiniMax H3 Max debuts #1 in Image-to-Video on Artificial Analysis

jfischoff · x · 2026-08-27

H3 Max, a post-trained version of MiniMax H3 developed by fal, debuts at #1 in Image-to-Video (with audio) on the Artificial Analysis Video Leaderboards, narrowly ahead of ByteDance's Dreamina Seedance 2.0 720p, and #3 in Text-to-Video — beating base MiniMax H3 on both. fal describes it as tuned for stronger prompt adherence and better aesthetics, co-optimized with their custom inference stack for higher throughput. It generates 5–15 second clips with native audio at up to 768p.

Related event: fal's post-trained H3 Max tops video leaderboard with order-of-magnitude speedup(12 posts)→

Original post →

More from Multimodal

Multimodal channel →