Advancing price-performance for developers with GPT‑5.6 in Kiro
The integration brings Sol, Terra, and Luna models to the development agent, reducing task costs on Terminal-Bench 2.1 by approximately 82%.
- Supports Sol, Terra, and Luna models across coding, planning, review, and testing workflows.
- Enables spec-driven development, checkpoint reviews before code implementation, and property-based testing.
- Developed in partnership with AWS to optimize runtime environments and model performance.
Developers using Kiro can now apply OpenAI's GPT-5.6 models directly to codebase-level software tasks with lower token costs.
