88Z AI, 12 June — Moonshot AI has released the open-source programming model Kimi K2.7 Code, which shows significant improvements over K2.6 in long-context processing, instruction following, and autonomous Agent execution. The model achieves an 11% to 31.5% increase in code benchmark scores and reduces token consumption by 30%, while maintaining the same API pricing as its predecessor. Enabling thinking mode is essential to unlock its optimal performance.
At a glance
- Code benchmark scores improved by 11%-31.5% with 30% less token usage
- Thinking mode required for best performance; API pricing matches K2.6
The update
Moonshot AI has officially open-sourced the new generation programming-specific model, Kimi K2.7 Code. Compared to the previous K2.6 version, the new model achieves significant enhancements in long-context programming scenarios, instruction following accuracy, and autonomous Agent execution capabilities. In various code benchmarks, Kimi K2.7 Code demonstrates performance gains ranging from 11% to 31.5%, while optimization efforts have reduced token consumption by 30%. Despite these substantial performance upgrades, the API pricing strategy remains consistent with K2.6, with no adjustments made. Officials specifically note that users must enable thinking mode to fully leverage the model's optimal performance, ensuring a superior programming assistance experience.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 月之暗面 KimiSource published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.