M2.7 is a mixture-of-experts model tuned for software engineering, agent harnesses and long-horizon office work; MiniMax reports a GDPval-AA Elo of 1495, which it calls the highest among open-source models (announcement). Its place in this corpus is the lab’s claim that it is the first model deeply participating in its own evolution: M2.7 wrote and repaired the harness used in its own reinforcement-learning experiments, then improved the learning process from the results. It succeeds M2.1 in the M2 line.
AISystem
MiniMax M2.7
MiniMax's M2-series coding and agent model, presented by its lab as the first of its models to take part in its own evolution by building and refining its own training harness.
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to March 23, 2026 (theinnermostloop.substack.com)
Development timeline
Newest reporting dates first. Reporting dates may differ from event dates.
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.