MiniMax’s H3 multimodal model, released July 31 with open weights, drives the 86.5% market-implied odds through consistent top rankings on image-to-video leaderboards including LMArena and Artificial Analysis audio-video evaluations. The system generates up to 15-second 2K clips with native stereo audio from image or reference inputs, excelling in motion transfer, prompt adherence, and speed—often under 25 seconds median in production tests—while undercutting rivals on price-performance. Competitors such as Google Veo 3.1, ByteDance Seedance, and Kling 3.0 trail in head-to-head preference studies and lack equivalent open-ecosystem momentum as of early October. With the October resolution window closing soon, any late competitor updates would need substantial benchmark gains to shift trader consensus.