tl;dr: Moonshot AI released Kimi K2.7-Code, an open-source 1T-parameter MoE coding model (32B activated) with 256K context, native INT4 quantization, and roughly 30% lower thinking-token usage than its K2.6 predecessor. It improves on long-horizon agentic coding benchmarks but still trails GPT-5.5 and Claude Opus 4.8 on most evaluations (e.g., 62.0 vs 69.0/67.4 on Kimi Code Bench v2). Weights are under a Modified MIT license and it runs on vLLM, SGLang, and KTransformers.