Moonshot AI open-sources Kimi K2.7-Code
Moonshot AI has open-sourced Kimi K2.7-Code, a 1.1-trillion parameter Mixture-of-Experts coding model that cuts reasoning token usage by 30% while improving benchmark performance. The model is released under a Modified MIT License on Hugging Face and is also accessible via the Kimi API.
The release of Kimi K2.7-Code demonstrates a key trend in open-source AI: optimizing for operational token efficiency in complex coding tasks rather than just raw model capacity. By slashing reasoning-token usage by 30% while achieving double-digit gains on benchmarks, Moonshot AI makes high-quality agentic coding workflows far more economically viable.
* **Token Efficiency Focus:** A 30% reduction in reasoning tokens directly addresses the high costs of multi-step agentic workflows.
* **Agentic Optimization:** Stronger performance on benchmarks like MLS Bench Lite indicates better stability and success in long-horizon software engineering tasks.
* **Open-Source Strategy:** Releasing weights under a Modified MIT License on Hugging Face lowers the barrier for community adoption and fine-tuning.
DISCOVERED
1h ago
2026-06-12
PUBLISHED
3h ago
2026-06-12
RELEVANCE
AUTHOR
nekofneko