Kimi Code K2.7 (twitter.com)

Kimi has announced the release of Kimi Code K2.7, an open-source coding model that boasts significant improvements over its predecessor, K2.6. The model demonstrates enhanced performance metrics, achieving a 21.8% improvement on the Kimi Code Bench v2, 11.0% on the Program Bench, and an impressive 31.5% on the MLS Bench Lite. Notably, K2.7 achieves greater reasoning efficiency, with a 30% reduction in overthinking, which is expected to streamline coding tasks significantly. This release is significant for the AI/ML community as it not only offers a more capable tool for developers but also enhances long-horizon coding capabilities, leading to better instruction following and higher success rates in end-to-end coding tasks. Furthermore, a high-speed mode is set to launch soon, promising even faster performance. Developers can access Kimi Code K2.7 today via the Kimi API and their coding platform, positioning it as a competitive option in the evolving landscape of AI-based coding solutions.