MiniMax M2.7 was benchmarked against Claude Opus 4.6 across three TypeScript coding tasks run inside the Kilo Code AI coding assistant: building a full-stack event processing system, debugging from production logs, and conducting a security audit. Both models found all 6 bugs and all 10 security vulnerabilities. Claude Opus 4.6 produced more thorough fixes, 41 integration tests vs 20 unit tests, and better security implementations (e.g., scrypt vs SHA-256 for password hashing). MiniMax M2.7 delivered roughly 90% of the quality at 7% of the cost ($0.27 vs $3.67 total), making it a strong option for cost-sensitive use cases where detection matters more than fix depth.