Kimi K2.7 Code is Moonshot AI's open-source, coding-focused agentic model designed for long-horizon software engineering workflows. It is positioned as a successor to K2.6, delivering stronger coding and agent performance with substantial gains on real-world coding tasks that translate into higher end-to-end task success rates across complex engineering work. Beyond raw capability, the model is engineered to be more efficient, reducing thinking-token usage by roughly 30% compared with K2.6, which helps keep multi-step agentic runs economical and responsive over extended sessions.

Moonshot AI's own benchmark write-up evaluates K2.7 Code against K2.6, GPT-5.5, and Claude Opus 4.8 across six coding and agentic benchmarks. On coding suites, K2.7 Code posts meaningful improvements over K2.6, and those coding gains flow through to stronger autonomous-agent results. The release targets developers and teams building production coding assistants and tool-using agents who want an open-weight model with explicit agentic emphasis rather than a general-purpose chat model.