MiniMax launched the general-purpose multimodal model H3, unifying text, image, video, and audio generation. It supports native dual-channel audio-video output up to 15 seconds at 2K resolution, with commercial-level stability in advertising, e-commerce, gaming, and UI design. Its per-second generation cost is less than one-third of mainstream models at 2K. MiniMax will open-source H3 weights soon, promoting domestic chip adaptation and reducing barriers for multimodal AI applications.