On June 12, Moonshot AI's Kimi announced the release and open-sourcing of its Kimi K2.7 Code programming model. According to official reports, both internal and external benchmark tests reveal that, in comparison to the K2.6 model, the K2.7 Code exhibits markedly enhanced instruction-following abilities in long-context programming scenarios. It also shows improved performance in handling long-range programming tasks, along with a significant decrease in overthinking tendencies during such tasks, resulting in an average token consumption reduction of 30%. In terms of code capability evaluations, the Kimi Code Bench v2 saw a 21.8% improvement, Program-Bench an 11% boost, and MLS Bench Lite a remarkable 31.5% increase. When it comes to Agent autonomous execution capability evaluations, there was an approximate 10% performance enhancement. Enterprises and developers can access the K2.7 Code model via the Kimi API open platform. The standard input and output pricing for 1M tokens remains unchanged from the K2.6 model, while the cached input price has been adjusted to 1.3 yuan. Moreover, a high-speed version is set to be launched on June 15, offering an output speed 5-6 times faster than the standard version, albeit at twice the standard rate.