MiniMax于7月31日正式发布新一代多模态生成模型H3并宣布近期开源[1]。受此消息驱动，MiniMax-W（HK：00100）的股价盘中一度拉升15%[1]。
根据Artificial Analysis的评测，H3在视频编辑能力上排名全球第一[1]。在定价策略上，该公司大幅下探成本：H3视频生成定价为0.8元/秒（2K分辨率），而低分辨率版本（768p）更是低至0.1元/秒[1]，这一价格仅为业内同类旗舰模型的三分之一[1]。
作为MiniMax首款开源多模态生成模型，H3适配国产AI芯片[1]。
MiniMax announced the release of its next-generation multimodal generation model H3 on July 31, with the company's stock price rising as much as 15% during intraday trading[1]. The H3 model ranks first globally in video editing capabilities according to Artificial Analysis assessments[1], marking a significant advancement in the company's commercial capabilities.
The pricing strategy represents a competitive advantage in the market. H3 is priced at 0.8 yuan per second for standard video generation at 2K resolution, with lower-resolution 768p output available at 0.1 yuan per second—approximately one-third the cost of comparable flagship models from competitors[1]. This pricing reflects the transition from exploratory creative applications to production-ready commercial deployment. MiniMax also announced that H3 will be released as its first open-source multimodal generation model, with compatibility designed for domestic AI chipsets[1].
