Kimi K2.7-Code: open-source coding model with better token efficiency
Summary
Kimi K2.7 Code is a coding-focused agentic model built on K2.6, delivering improved token efficiency (~30% reduction in thinking tokens) and a MoE architecture with 1T total parameters. The model is evaluated against K2.6, GPT-5.5, and Claude Opus 4.8 across coding and agentic benchmarks, supports native INT4 quantization, and offers multiple deployment options (vLLM, SGLang, KTransformers). The page provides deployment guidance, usage examples, and licensing details (Modified MIT).