Kimi K2.7 Code
moonshotai/Kimi-K2.7-Code
Open Source · chat · open-weights
Intelligence
Context
262.1K
Max output
—
Input $/1M
$0.72
Output $/1M
$3.50
Modalities
text · image
Released
11 Jun 2026
Intelligence Index via Artificial Analysis · 0–100, higher is better
License: other · moonshotai/Kimi-K2.7-Code
AI summary
● machine-written
Moonshot AI releases Kimi K2.7 Code, open-weight coding model
Kimi K2.7 Code is an open-source, coding-focused mixture-of-experts model with 1 trillion total parameters and 32 billion active parameters, designed for long-horizon software engineering tasks. The model supports text and image input, operates with a 256K context window, and includes thinking-mode reasoning that reduces token usage by approximately 30% compared to its predecessor K2.6.
What's new
- Achieves +21.8% on Kimi Code Bench v2 over K2.6 (62.0 vs 50.9)
- Reduces thinking-token usage by ~30% compared to K2.6
- Supports 256K context window with multi-turn thinking preservation
- Model weights open-sourced on Hugging Face
Best for
Long-horizon software engineering and multi-file refactoringAutonomous coding task execution and agentic workflowsInteractive coding sessions requiring extended contextComplex program comprehension and debugging