Moonshot ships Kimi K2.7 Code on June 13 — coding-specialized variant cuts thinking tokens by ~30% vs K2.6, directly lowers cost of long-agent-run coding workflows
Moonshot's Kimi K2.7 Code (released June 13) is a coding-specialized variant that cuts thinking tokens by approximately 30% versus the K2.6 baseline. The token-efficiency improvement directly lowers the cost of long-agent-run coding workflows — the operational economics dimension that distinguishes production-deployment-viable coding agents from research-tier ones.
The substantive piece is the production-economics dimension as a coding-agent competitive axis. Pre-K2.7-Code coding-agent procurement evaluation focused primarily on capability (SWE-Bench scores, Terminal-Bench performance) with cost as a secondary consideration. The 30% thinking-token reduction makes operational economics a first-class evaluation dimension — the same workload at 30% lower inference cost has direct procurement-deployment economics implications.
The competitive read against GLM-5.2's 6.8x-cheaper-than-GPT-5.5 economics on coding benchmarks is that the Chinese-open-weight cost-leadership pattern is sustaining across multiple vendors. Kimi K2.7 Code's 30% reduction over K2.6 plus GLM-5.2's structural cost advantage together represent an open-weight cost-pressure on closed-source vendors that didn't exist in H1 2026.
LLM Stats — AI Updates Today (June 2026) → · Hugging Face — Best Open-Source LLM Models in 2026 →