MiniMax-M3 debuts, eclipsing GPT-5.5 and Gemini 3.1 Pro on key benchmark performance for just 5-10% of the cost

Chinese AI startup MiniMax released M3, a multimodal large language model with a 1-million-token context window and claimed frontier-level coding and agentic performance. The company says M3 beats select OpenAI and Google models on key benchmarks while costing roughly 5% to 20% as much, and plans to release open weights within about 10 days.

M3 reflects a broader push by Chinese AI labs to compete on both performance and price, pressuring proprietary model providers to justify premium API costs. If the open-weight release delivers on enterprise customization and benchmark claims, it could accelerate adoption of lower-cost frontier models for coding, agents, and long-context workflows.