Kimi K2.7-Code efficiency gains questioned by practitioners

Moonshot AI has released Kimi K2.7-Code, an update to its K2 coding model, claiming a 30% reduction in thinking tokens and improved performance. This update is designed to lower inference costs for agentic workflows by optimizing code generation. However, independent practitioners and researchers are questioning the validity of Moonshot AI's proprietary benchmarks, suggesting the claimed efficiency gains may not hold up under external scrutiny. Early tests indicate that while the model may be more direct in its approach, its overall capability and reliability on certain tasks have not demonstrably improved, and in some cases, have regressed compared to its predecessor.