The Chinese-made AI model 'Kimi K2.7 Code' has been released as an open model, boasting the best coding performance in the Kimi series and low token consumption.
Chinese company
Kimi K2.7 Code: Open-Source Agentic Coding Model
https://www.kimi.com/resources/kimi-k2-7-code
🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!
— Kimi.ai (@Kimi_Moonshot) June 12, 2026
🔷 Improved coding & agent performance over K2.6: +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, and +31.5% on MLS Bench Lite.
🔷 Reasoning efficiency: Less overthinking, with 30% lower… pic.twitter.com/jFS7I40avs
Kimi K2.7 Code is a MoE model with a total of 1 trillion parameters and 32 billion active parameters, and a maximum context length of 262,144 (256K) tokens.
The benchmark results for 'Kimi K2.7 Code,' 'Kimi K2.6,' 'GPT-5.5(xhigh),' and 'Claude Opus 4.8(xhigh)' are shown below. The Kimi K2.7 Code shows a significant improvement in coding task performance compared to the previous generation model, the Kimi K2.6 .
AI agents can sometimes accumulate unnecessary token consumption to perform a single task. Kimi K2.7 Code has improved inference efficiency and can perform tasks with approximately 30% fewer thinking tokens compared to Kimi K2.6. The graph below shows the number of tokens consumed on the horizontal axis and performance on the vertical axis, demonstrating that Kimi K2.7 Code is both high-performing and highly efficient compared to Kimi K2.6.
The Kimi K2.7 Code is available as an open model at the following link.
moonshotai/Kimi-K2.7-Code · Hugging Face
https://huggingface.co/moonshotai/Kimi-K2.7-Code
Related Posts:
in AI, Posted by log1o_hf