MiniMax M2.7 is a 229B-parameter large language model in MiniMax's M2 series, designed for coding, agentic workflows, and professional productivity. Hosted on Ollama's library under the minimax-m2.7 key, it runs through the standard local chat interface and is wired into developer tooling such as Claude Code, OpenCode, and Hermes Agent, making it approachable for teams that already use those harnesses. The accompanying Together AI listing frames the same model as production-scale software engineering work, emphasizing long-horizon agentic execution and native Agent Teams that coordinate multiple roles with persistent identity and autonomous decision-making.

Reported evaluations on Together AI place M2.7 near the top of its peer group on real-world coding and machine learning engineering tasks, with a 56.22% score on SWE-Pro that matches GPT-5.3-Codex, a 55.6% result on VIBE-Pro for end-to-end project delivery across Web, Android, and iOS, and a 66.6% medal rate on MLE Bench Lite across twenty-two machine learning competitions. The provider also describes a self-directed optimization loop in which an internal version of the model ran more than one hundred rounds of failure analysis and code modification, reportedly yielding a thirty-percent gain on internal programming benchmarks. With a 200K-token context window, tool calling, and open weights, M2.7 is a strong fit for developers who want a locally deployable model aimed at multi-step engineering and multi-agent pipelines rather than short conversational exchanges.