MiniMax-M2.7 Is Now Available in Kilo. Here’s How It Performs.

MiniMax-M2.7 is now available in Kilo's IDE extension, CLI, and Cloud Agents. Benchmark results show it scores 86.2% on PinchBench (5th out of 50 models), placing it near Claude Opus 4.6 and GPT-5.4. On Kilo Bench's 89-task autonomous coding evaluation, it passes 47% of tasks, finishing second overall. A key behavioral trait is its exploration-heavy approach — reading extensive context before acting — which helps on complex reasoning tasks but can cause timeouts on time-sensitive ones. It consumes ~2.8M input tokens per trial on average, the highest of any tested model. Priced at $0.30/M input and $1.20/M output, it's significantly cheaper than frontier models with comparable benchmark scores. The post recommends M2.7 for deep refactors and codebase-wide changes, while suggesting lighter models for fast iteration cycles.