MiniMax's recent H3 multimodal model release and its accelerated H3 Max variant have driven the 76.5% market-implied odds, with top Elo rankings on Artificial Analysis image-to-video leaderboards as of early October 2026. The July 31 launch introduced unified text-image-video-audio processing, native stereo sound, up to 2K resolution, and strong prompt adherence plus motion consistency for image-to-video tasks, outperforming closed competitors in speed and reference control while offering open weights for community fine-tuning. Post-training optimizations by platforms like fal have further boosted throughput without quality loss, solidifying trader consensus around MiniMax ahead of end-of-October resolution. Google and ByteDance models trail due to narrower leads in recent human preference evaluations, with limited near-term catalysts expected to shift the positioning.