Global Markets
Home›Global Markets›China›MiniMax launches H3 open-weight video model, targeting…
MiniMax launches H3 open-weight video model, targeting video creators
The new model, available via an API and the Hailuo AI platform, is set for full weight release in the coming days.
Chinese AI firm MiniMax has launched H3, a multimodal video generation model aimed at commercial content creation across film, advertising, video games, and product design. The company is positioning H3 as a pushback against closed-source “dominance,” saying it will be offered with open weights and competitive pricing, while rival ByteDance has rolled out its Seedance 2.5 model.
According to benchmark platform Artificial Analysis, H3 was the world’s most powerful AI model in video editing at the time of launch. However, the model lagged Google’s Gemini Omni Flash in text-to-video tasks, and it ranked behind both ByteDance’s Seedance 2.0 and Gemini Omni Flash in image-to-video capabilities.
MiniMax said H3 is currently available through an application programming interface and its consumer-facing Hailuo AI platform, with full model weights planned for release in the coming days. The company said H3’s weights will be released under the MiniMax Community License, which allows free non-commercial use and commercial use for organizations with under US$20 million in annual revenue, with attribution requirements.
MiniMax also said it previously kept its Hailuo video models proprietary, but chose to open H3’s weights to support the open-source community and allow users to customize versions using a broad range of AI hardware. In its blog post, the company argued that closed-source models have slower iteration and a less open ecosystem than large language model fields.