MiniMax releases M3.1 Flash, reducing agent task costs by 70% compared with the previous generation.

PANews October 20 news, according to the MiniMax official blog, MiniMax has officially released its new-generation model M3.1 Flash. The cost of using it for agent tasks is 70% lower than that of the previous generation, online inference speed is nearly 3 times faster, and output speed exceeds 200 tokens per second; API pricing is $0.15 per million input tokens and $0.6 per million output tokens. The model scored 61.7 on the OSWorld V2 benchmark, ranking first among domestic models. The company said that the anonymous model "Space Bunny," which previously processed over 200 trillion tokens, was an early version of M3.1 Flash. The model is now available through the MiniMax API, M Plan, and MiniMax Code, and the model weights will be open-sourced later. MiniMax launched M3.1-Flash-Preview in September.