MiniMax H3: First Open-Weights Model Tops the Global AI Video Rankings

MiniMax H3, a 33-billion-parameter open model from Chinese company MiniMax, launched on August 3 and immediately claimed first place in the global AI video editing leaderboard and second in text-to-video generation, according to Artificial Analysis. The model generates 4–15 second clips with stereo audio and accepts up to nine reference images in a single request. It is available for local deployment, and commercial use is open to companies with up to $20 million in annual revenue.

Until now, the top of AI video leaderboards was held exclusively by closed services — Sora, Runway, and Kling. Open weights change the equation: the model can be fine-tuned on your own footage without paying for an API, shifting control over AI video generation away from large platforms and toward whoever holds the data.