OpenAI has made its GPT-5.6 model family available inside Kiro, a software development agent built by AWS that structures AI-native coding around specifications, designs, and executable tasks. The integration brings three model tiers — Sol, Terra, and Luna — into the workflows where teams plan, build, review, and test software.
Kiro turns high-level intent into structured implementation plans, technical designs, and task context. That structured grounding helps GPT-5.6 understand what a team is building, how the system should work, and what the final implementation needs to accomplish. Developers can apply the models to long-running development work anchored in their requirements, codebase, and team standards.
What's new
- Three tiers: Sol (flagship capability), Terra (balanced everyday work), and Luna (fastest, most affordable).
- Spec-driven workflow: Kiro grounds the model in clear requirements and designs from the start, reducing missteps and wasted iterations.
- Checkpoint reviews: Developers review and refine model output at key stages before changes are implemented.
- Property-based testing: Correctness of implementation is checked automatically.
- Cost reduction: On Terminal-Bench 2.1, GPT-5.6 Terra completed successful tasks in Kiro at roughly 82% lower cost.
Pricing and availability
OpenAI also announced price cuts for the API that took effect July 30, 2026. GPT-5.6 Terra now costs $2 per million input tokens and $12 per million output tokens (20% less). GPT-5.6 Luna costs $0.20 per million input tokens and $1.20 per million output tokens (80% less). Sol pricing remains unchanged. In ChatGPT Work and Codex, Free and Go users can access Terra, while Plus, Pro, Business, and Enterprise users can choose Terra and Luna. Subscription prices and quota budgets stay the same, but Terra and Luna usage now consumes fewer credits.
OpenAI also introduced Fast mode for GPT-5.6 Sol in the API, replacing the previous Priority Processing tier. Fast mode delivers up to 2.5× faster speeds than Standard processing at twice the price, with no change in intelligence. Requests tagged "priority" automatically use Fast mode. Pricing changes will begin rolling out in AWS later the same day.
Why it matters
The integration signals a shift toward matching model capability to task stakes inside a single development workflow. A team can use Sol for planning and uncertainty resolution, then hand off well-specified implementation, testing, and evaluation to Luna or Terra — paying only for the intelligence each step actually requires. OpenAI notes that Luna delivers performance comparable to frontier-class models from a year ago at roughly 6 cents on the dollar per task and nearly nine times the speed, while outperforming Fable 5 on Agents' Last Exam at an estimated 99% lower cost per task.
Our take
The real leverage here isn't just cheaper tokens — it's the spec-driven harness that lets teams route work to the right tier without losing context. If Kiro's structured approach reliably cuts the 82% cost figure on real projects, the economics of agentic coding shift from "experiment budget" to "default workflow" for a much wider set of teams.
