Moonshot AI Releases Kimi K2.7-Code with Efficiency Gains

Moonshot AI has released Kimi K2.7-Code, an open-source update to its K2 coding model family, boasting leaner reasoning and double-digit performance gains. The model is built on the same trillion-parameter mixture-of-experts architecture as its predecessor K2.6 and drops in via an OpenAI-compatible API. Moonshot AI claims K2.7-Code reduces thinking-token usage by 30% compared to K2.6, which could directly affect inference costs for teams running agentic workflows. However, practitioners have raised questions about the model's performance on independent benchmarks. Researcher Elliot Arledge ran K2.7-Code against K2.6 and Claude Fable 5 on KernelBench-Hard, a public benchmark, and found that K2.7-Code produced real authored Triton kernels but regressed in some areas. Developer Sugumaran Balasubramaniyan also challenged Moonshot AI's benchmark choices, noting that every model 'improves' on its own test suite.