OpenAI Developers said they worked with AWS to optimize GPT-5.6 inside Kiro. In Kiro's spec-driven environment, GPT-5.6 Terra delivered about an 82% cost reduction per successful Terminal-Bench 2.1 task. The pairing brings Sol, Terra, and Luna into AWS's requirements-to-implementation workflow with property-based testing gates, targeting long-running software work rather than one-shot completion.
Key Takeaways
- βGPT-5.6 Sol, Terra, and Luna now power AWS Kiro's spec-driven coding workflow.
- βOpenAI reports ~82% lower cost per successful Terminal-Bench 2.1 task for Terra in Kiro.
- βKiro pairs specs, review checkpoints, and property-based tests for long-running agents.