Updated
Updated · OpenAI · Aug 24
OpenAI Brings GPT-5.6 to Kiro, Cutting Terminal-Bench 2.1 Costs by 82%
Updated
Updated · OpenAI · Aug 24

OpenAI Brings GPT-5.6 to Kiro, Cutting Terminal-Bench 2.1 Costs by 82%

3 articles · Updated · OpenAI · Aug 24

Summary

  • GPT-5.6 is now available in Kiro, adding OpenAI’s Sol, Terra and Luna models to workflows for planning, building, reviewing and testing software.
  • Kiro’s spec-driven setup feeds the models requirements, technical designs, codebase context and team standards, aiming to deliver higher-quality code with fewer iterations on long-running tasks.
  • Terminal-Bench 2.1 testing found GPT-5.6 Terra completed successful tasks in Kiro at roughly 82% lower cost, underscoring OpenAI’s pitch of stronger performance per dollar.
  • OpenAI and AWS said they have optimized both Kiro and the models already and will keep working together to improve performance across the software development lifecycle.

Insights

With GPT-5.6 automating the coding lifecycle, are human developers destined to become mere reviewers of AI-generated software?
Could Kiro's hyper-efficient AI actually create a massive hidden technical debt that human engineers can no longer decipher?
How secure is your proprietary codebase when an autonomous AI agent has full access to query enterprise knowledge bases?