OpenAI Adds GPT-5.6 Model Family to AWS Kiro
OpenAI has made its GPT-5.6 model family available in Kiro, AWS’s software development agent. The release brings GPT-5.6 Sol, Terra and Luna into workflows for planning, building, reviewing and testing software. OpenAI and AWS said testing on Terminal-Bench 2.1 found that GPT-5.6 Terra completed successful tasks in Kiro at roughly 82% lower cost.
The companies position the integration around price-performance for development teams. GPT-5.6 is designed to deliver more useful work per token, stronger performance per dollar and on-demand capability for complex tasks. Within Kiro, the models can be used for long-running work informed by requirements, a codebase and team standards.
Structured context for AI-assisted development
Kiro turns high-level intent into requirements, technical designs and executable tasks. That structured context is intended to give GPT-5.6 a clearer view of the software being built, the expected system behaviour and the implementation outcome required by a team.
OpenAI said developers can use GPT-5.6 in Kiro to convert product ideas and requirements into structured implementation plans, complete complex multi-step coding tasks and work with context drawn from their codebase and established standards. The platform also supports spec-driven development, allowing developers to review and refine model output at key checkpoints before changes are implemented.
The workflow includes property-based testing as a way to check implementation correctness. Rather than treating a generated change as a final result, the approach combines specifications, task context, review points and testing within the same development process.
Cost result and model options
The roughly 82% cost reduction reported for GPT-5.6 Terra refers to successful tasks in Kiro on Terminal-Bench 2.1. OpenAI and AWS said they worked together to optimise both the Kiro environment and OpenAI models. The companies attribute the result to Kiro grounding the model in requirements, technical designs and task context from the beginning, helping it reach working solutions with fewer missteps.
Swami Sivasubramanian, Vice President of Agentic AI at AWS, said the addition expands developers’ options for accelerating AI-native development through Kiro. Colleen Kapase, Vice President of Strategic Global Partnerships and Ecosystems at OpenAI, said the GPT-5.6 family gives teams more scope to match intelligence, speed and cost to stages of the software development lifecycle.
OpenAI and AWS said they will continue improving OpenAI model performance in Kiro. For businesses, the immediate implication is to assess GPT-5.6 models in Kiro against their own requirements, repository context, review controls and testing practices before applying the workflow to complex engineering tasks.