MiniMax H3: First Open-Weight Video Model With Native Stereo Audio in a Single Pass

MiniMax released H3 on August 3, 2026 — its first open-weight video model. H3 generates up to 2K video with native stereo audio in a single forward pass, eliminating separate audio generation and synchronization. Weights are on Hugging Face (MiniMaxAI/MiniMax-H3). The omni-modal model accepts text, images, video, or audio and produces clips up to 15 seconds.