Moonshot AI has officially released Kimi K2.7-Code, an open-weights coding model optimized for long-horizon software engineering and cost-efficient agentic reasoning.
Moonshot AI has officially released Kimi K2.7-Code, an open-weights Mixture-of-Experts coding model featuring 1 trillion parameters and a 256K context window. Optimized for long-horizon software engineering tasks like codebase-wide refactoring and debugging, the model achieves a 30% reduction in reasoning-token usage compared to its predecessor. Kimi K2.7-Code supports multimodal inputs, runs in a dedicated reasoning-heavy thinking mode, and is available for developers via Hugging Face, Ollama, and the Kimi API.
Moonshot AI is directly challenging proprietary coding leaders by offering a massive open-weights alternative that significantly optimizes reasoning cost.
- –**Cost-Efficiency Focus:** A 30% reduction in reasoning tokens is a major win for developer teams running expensive agentic loops.
- –**Multimodal Capabilities:** Support for image and video inputs makes Kimi K2.7-Code highly competitive for UI-driven or complex visually-guided software engineering tasks.
- –**Always-on Thinking:** The model runs strictly in a reasoning-heavy thinking mode to prioritize accuracy over raw speed.
DISCOVERED
3d ago
2026-06-13
PUBLISHED
3d ago
2026-06-13
RELEVANCE
AUTHOR
AiquestAcademy