Asia

MiniMax says its M2.7 model helps develop itself

Internal agent loops automate up to half of training workflow, capability marketing arrives before independent verification

Images

Image description Image description the-decoder.com
MiniMax M2.7 compared with Sonnet 4.6, Opus 4.6, Gemini 3.1 Pro, and GPT-5.4 across eight benchmarks. M2.7 scores close to the leading proprietary models in most tests. | Image: MiniMax MiniMax M2.7 compared with Sonnet 4.6, Opus 4.6, Gemini 3.1 Pro, and GPT-5.4 across eight benchmarks. M2.7 scores close to the leading proprietary models in most tests. | Image: MiniMax MiniMax
How M2.7 develops itself: researchers set goals and guidelines, then the AI agent takes over large parts of the development process on its own. The example workflow below shows how experiment planning, code changes, and evaluation feed into each other. | Image: MiniMax How M2.7 develops itself: researchers set goals and guidelines, then the AI agent takes over large parts of the development process on its own. The example workflow below shows how experiment planning, code changes, and evaluation feed into each other. | Image: MiniMax MiniMax

MiniMax claims M2.7 model helped develop itself, autonomous optimization loops shift AI work from engineers to internal agents, China pushes capability narratives while export controls and standards battles loom