Kimi K2.7-Code: open-source coding model with better token efficiency (huggingface.co) AI

Moonshot AI’s Kimi K2.7-Code is an open-source, coding-focused agentic model built on Kimi K2.6, claiming improved token efficiency (about 30% fewer thinking tokens) and better performance on long-horizon software engineering tasks. The Hugging Face model card includes deployment and usage examples via Transformers, vLLM, and SGLang, along with reported benchmark results and specifications such as a 256K context length and native INT4 quantization.