Kimi K2.7 Code
Coding-only K2.6 derivative with better instruction-following in long contexts and about 30% less overthinking; thinking mode always on.
Specialised for long-horizon coding agents inside Kimi Code. Moonshot claims roughly 30% fewer reasoning tokens than K2.6 and about 10% better agentic performance, with a 'HighSpeed' variant at about 180 tokens/s. No non-thinking mode.
- Date
- Friday, 12 June 2026
- Lab
- Moonshot AI
- Kind
- open-weights
- Access
- open weights
Figures
| Measure | Value | Measured by |
|---|---|---|
| Overthinking reduction vs K2.6 | ~30% average, company claim | company |
| Context window | 256K | company |
Date from Epoch AI (2026-06-12) and HF README commit on that date. License is modified MIT per Epoch/HF.
Sources
- platform.kimi.ai/docs/guide/kimi-k2-7-code-quickstart
- huggingface.co/moonshotai/Kimi-K2.7-Code
- epoch.ai/models/kimi-k2-7-code
- gigazine.net/gsc_news/en/20260615-kimi-k2-7-code/
This record was checked against its sources on 6 October 2026. How we check