China's MiniMax launches updated AI model M3.1 Flash with faster speeds, lower costs

Key facts

• M3.1-Flash-Preview is the default model in MiniMax Code, the company's coding-agent environment.
• The model is positioned for fast, reliable everyday software development, including bug fixes and feature work.
• MiniMax has not published standalone M3.1 benchmark suites, public general API model IDs, or separate API pricing.

Chinese AI startup MiniMax launched an updated version of its AI model, M3.1 Flash, on Monday. The company stated the updated model delivers "generational" improvements in areas such as agentic tasks, programming, and cybersecurity. MiniMax also claimed its online inference speed has nearly tripled while usage costs have been reduced by 70%. The M3.1-Flash-Preview model is integrated into MiniMax Code, the company's coding-agent environment, and is positioned for fast, reliable everyday software development. However, MiniMax has not published standalone benchmark suites, public general API model identifiers, or separate API pricing for the model.