The latest Kimi K2.7 Code programming model is open sourced on the dark side of the moon, and the high-speed version will be launched next Monday According to Beating monitoring, Moonshot AI has released and open-source the latest Kimi K2.7 Code programming model. In long context programming scenarios, Kimi K2.7 Code improves the performance of instruction adherence and long-range programming tasks, while reducing the tendency to overthink, resulting in a 30% reduction in average token consumption. In the Kimi Code Bench v2 benchmark test evaluating code capability, the new model showed a 21.8% performance improvement, and in benchmark tests such as Kimi Claw 24/7 Bench evaluating the autonomous execution ability of intelligent agents, the performance improvement was about 10%. Starting today, enterprises and developers can call Kimi K2.7 Code through the Kimi API open platform. The standard input and output prices for each 1M token are 6.5 yuan and 27 yuan, respectively, which is the same as the K2.6 model. The input price for cache hit is adjusted to 1.3 yuan. The default model of Kimi Code Plan has been synchronously upgraded, and using Kimi K2.7 Code requires enabling the thinking mode to achieve optimal performance. Moonshot AI also announced that it will launch a high-speed version model with an output speed increase of 5-6 times on June 15th. The output speed in regular scenarios is about 180 Tokens/s, and the call price is twice that of the regular version. [Original link]