Kimi K2.7-Code: open-source coding model with token efficiency focus

Category: AI
Status: confirmed
Sources: HuggingFace model card, HN discussion
Summary: Moonshot AI released Kimi K2.7-Code, a coding-focused open-weight model available on HuggingFace under a Modified MIT license with attribution requirements. The model targets improved token efficiency for coding workloads compared to its K2.6 predecessor. Vendor-reported coding benchmark results place it below GPT-5.5 and Opus 4.8 but ahead of several mid-tier models on coding task geometric mean. Available as open weights for self-hosted deployment.
Why it matters: Open-weight coding models with permissive licenses provide a self-hosted alternative to API-only services for teams with data-residency or cost constraints.