MiniMax’s H3 model, released in late July 2026, drives the 86.5% market-implied odds through consistent top rankings on image-to-video leaderboards such as LMArena and Artificial Analysis. It leads in blind preference tests with strong scores around 1,495, offering native stereo audio, multi-modal reference inputs up to nine images plus video and audio clips, first- and last-frame control, and up to 2K resolution in 4–15 second clips. Post-training accelerations like fal’s H3 Max and community LoRAs have further boosted speed and accessibility without sacrificing core capabilities. With roughly two weeks until the October resolution, rivals including Google Veo 3.1, Kling 3.0, and Seedance 2.5 trail in direct I2V benchmarks, though ongoing optimizations could narrow gaps before month-end.