MiniMax's product blitz: new models for code, music, and video in one broad refresh

Chinese AI startup MiniMax has expanded its product and model portfolio with a slate of new offerings spanning language, video, speech, music, and code generation. The company, best known for its conversational AI platform and the Hailuo video generation tool, announced the updates on its website and product pages.

The centerpiece of the refresh is MiniMax M3, the latest version of the company's language model. MiniMax has not released detailed benchmark comparisons or parameter counts for M3, but the company positions it as a direct successor to MiniMax M2.7 and M2.5, with improvements in reasoning, generation quality, and performance across a broader range of tasks. The company's language models have historically been competitive in Chinese-language tasks and are used both by consumers through chat interfaces and by enterprise customers via the MiniMax API.

Alongside M3, MiniMax introduced MiniMax Hailuo 2.3 and 2.3 Fast, the latest iteration of its video generation model. The Fast variant appears to focus on inference speed, likely targeting creators who need rapid prototyping or iterative output. Hailuo has been a key differentiator for MiniMax in the Chinese AI market, where video generation competition includes players like Kuaishou's Kling and Alibaba's Tongyi Wanxiang.

In a notable expansion into new modalities, MiniMax launched MiniMax Code, a code-specific product that competes with established coding assistants such as GitHub Copilot, Amazon CodeWhisperer, and Chinese rivals like iFlytek's Spark Code. The move suggests MiniMax is betting on growing demand for AI-assisted software development tools, a space that has seen rapid adoption across enterprise and individual developer markets.

The company also updated its audio generation capabilities with MiniMax Speech 2.8 and MiniMax Music 2.6. These models are part of the company's broader push to offer multimodal generation, matching trends seen at OpenAI and Google DeepMind. The speech model likely improves naturalness and expressiveness, while the music model targets content creators, game developers, and social media producers.

On the product side, MiniMax continues to operate its consumer apps: the video generation platform Hailuo AI, the voice assistant Voice, and the social chatbot Xingye, which is popular among younger Chinese users for emotional companionship and role-playing. The company also maintains its API offering with a token-based pricing plan for developers and businesses.

MiniMax's research page and about us sections describe a mission of creating intelligence with everyone, positioning the company as a research-driven AI lab rather than merely a product company. Founded in 2021 by Yan Junjie, formerly a VP at SenseTime, and others, MiniMax has raised over $1 billion in funding, including a strategic investment from Alibaba in early 2024. The company is part of China's new AI wave that includes Zhipu AI, Baichuan, and Moonshot AI.

The timing of the update is significant. The Chinese AI sector is in a heated race for talent, capital, and user adoption. MiniMax's broad product refresh, spanning language, video, audio, and code, suggests a strategy of pursuing horizontal expansion to capture multiple market segments. This contrasts with more focused rivals: Moonshot AI concentrates on long-context language models, while Zhipu AI emphasizes enterprise and open-source offerings.

The MiniMax M3 language model and Hailuo 2.3 video model are available now through the company's consumer apps and API. Pricing details for the new models have not been fully disclosed, but the company's existing token-based plan starts at competitive rates for Chinese-language tasks.