MiniMax: MiniMax M2.7

MiniMax M2.7 is a text-only model with a 200K-token context window and a 128K-token output ceiling, making it practical for long-document tasks. It supports tool use and reasoning, which means it can handle multi-step agentic workflows and function-calling integrations. Structured output support is unconfirmed, so developers who depend on guaranteed JSON schemas should verify that independently before committing. At $0.25 per million input tokens and $1.00 per million output tokens, MiniMax M2.7 sits at the affordable end of reasoning-capable models. Its blended benchmark score of 64.8 covers only three benchmarks, so confidence in that number is limited. The coding subscore of 86.8 is notable and may attract developers building code-heavy pipelines, but the agentic score of 42.2 is low, meaning teams running complex autonomous agents should weigh that gap carefully before shortlisting it.