Editorial illustration for MiniMax M2.7 Agent Scores 56.22% SWE‑Pro, 57% Terminal Bench 2, ELO 1495
MiniMax M2.7: Open Source AI Coding Agent Breaks Records
MiniMax M2.7 Agent Scores 56.22% SWE‑Pro, 57% Terminal Bench 2, ELO 1495
MiniMax's new model gets good grades on the tough software engineering tests. But that's table stakes now. What matters is how it got them.
M2.7 ran a hundred rounds of self-directed optimization, boosting its performance by thirty percent with no hand-holding. It can read a stack of annual reports, cross-reference them with outside research, build a financial forecast, and spit out a polished PowerPoint. It does this with a ninety-seven percent success rate across forty complex jobs.
And as of now, the whole thing is open source. You can download the weights from Hugging Face and run it yourself.
On Terminal Bench 2 (57.0%) and NL2Repo (39.8%), both of which demand a high degree of system-level comprehension, MiniMax M2.7 performs solidly.
The benchmark scores are a distraction. A model that matches GPT-5.3-Codex on SWE-Pro and sits second only to a few premium closed models on the GDPval-AA leaderboard is impressive. But the real shift is that this capability is now a public utility.
Anyone can take this thing, which taught itself to be better, and point it at their own problems. The junior analyst workflow, the precise skill compliance, the autonomous optimization, these are no longer features locked behind a corporate API. They're files on a server.
The frontier just got a lot closer to the ground.
Related Reading
Open Source Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source SDOF Adds Two Defensive Layers via Intent Router and StateAwareDisp
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
From Archives 20‑year‑old arrested after Molotov attack on Sam Altman’s home, OpenAI threat
From Archives Google, Arm, and Qualcomm boost Gemma 4 AI to run 4× faster on Android
Common Questions Answered
What benchmark scores did the MiniMax M2.7 agent achieve?
The MiniMax M2.7 scored 56.22% on the SWE-Pro suite and 57% on Terminal Bench 2, demonstrating strong performance in code generation capabilities. These scores position it as a competitive open-source agent in technical evaluation metrics.
How does the MiniMax M2.7 rank in the GDPval-AA evaluation?
In the GDPval-AA evaluation, the MiniMax M2.7 achieved an ELO score of 1495, which is the highest among open-source models. This score places it second only to Opus 4.6, Sonnet 4.6, and GPT-5.4 in the overall ranking.
Where can developers access the MiniMax M2.7 model?
The MiniMax M2.7 model is now publicly available on Hugging Face, with its weights released for the first time. This open release allows developers to explore and potentially utilize the model's advanced capabilities.
Further Reading
- MiniMax M2.7: The Self-Evolving AI Model That Rivals Claude and ... — Wavespeed.ai
- MiniMax M2.7 - API Pricing & Providers - OpenRouter — OpenRouter
- MiniMax-M2, a model built for Max coding & agentic workflows. — GitHub - MiniMax-AI
- MiniMax-M2.7 - Intelligence, Performance & Price Analysis — Artificial Analysis
