Chinese startup MiniMax released the H3 multimodal video generation model on July 31, capable of producing 2K resolution videos with stereo sound at less than one-third the cost per second of mainstream competitors. The model accepts text, images, videos, and audio simultaneously, with primarily commercial applications in advertising, e-commerce, and video gaming. MiniMax plans to release model weights as open source in early August, allowing anyone to run it locally for free. Founded in December 2021 in Shanghai, the company raised approximately $619 million in its Hong Kong IPO in January 2026, with backing from Alibaba and Tencent. This price reduction and open-source commitment challenge the viability of decentralized AI networks and the AI-crypto thesis built on high compute costs.
Source: Read the original article
