Kiro adds GPT-5.6 model family

This is less about adding another model dropdown and more about making frontier-model economics useful across the software lifecycle.

• –Sol, Terra, and Luna create clear tradeoffs between capability, throughput, and cost
• –Kiro grounds agents in requirements, technical designs, team standards, and executable tasks
• –Property-based testing and review checkpoints address the reliability gap in autonomous coding
• –Multi-provider access reduces lock-in, but makes model routing and credit governance more important
• –OpenAI’s reported Terminal-Bench cost savings are promising, though developers should validate quality and completion rates on their own codebases