MiniMax debuts AI model built for long and complex coding tasks
Shanghai-based company says M3 can process data five times faster than its predecessor, while also slashing inference costs
Chinese artificial intelligence start-up MiniMax has unveiled its latest flagship AI model, M3, designed to anchor the company’s push into coding agents and automated workflows.
The Shanghai-based company said on Monday that the model’s redesigned architecture reduced computational requirements to as little as one-twentieth of previous levels, slashing inference costs while boosting response speeds.
Notably, MiniMax said M3 could process up to 1 million tokens of data at once – five times more than its predecessor, the M2.7. This allowed the model to handle long, complex programming projects. In one benchmark test cited by the company, M3 successfully figured out how to optimise software running on Nvidia’s Hopper chips.
According to the company’s WeChat post, MiniMax-M3 beat rival models like OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro on the major coding benchmark SWE-Bench Pro, showing that the model was highly capable at handling software engineering and automated tasks.
The company did not disclose the size of the model or the computing infrastructure it used for training.