GPT‑5.6 lands in AWS's Kiro agent
OpenAI has made its GPT‑5.6 models available in AWS's Kiro coding agent, and says joint optimization work cut the cost of completed Terminal-Bench 2.1 tasks by roughly 82% — a vendor figure with no stated baseline.
OpenAI said its GPT‑5.6 model family — including the variants it calls Sol, Terra, and Luna — is now available in Kiro, the AWS software development agent built around spec-driven development, in which high-level intent is converted into requirements, technical designs, and executable tasks before code is written.
The companies also say they worked together to optimize the Kiro environment and OpenAI's models for each other. OpenAI cites one figure from that work: in testing on Terminal-Bench 2.1, GPT‑5.6 Terra completed successful tasks in Kiro at "roughly 82% cost reduction." The post does not state the baseline that reduction is measured against — an earlier model, a different harness, or a different configuration — nor does it publish task-success rates, token counts, or absolute dollar figures alongside it. Treat the number as a vendor claim pending independent replication.
Beyond that, the announcement is availability copy: OpenAI lists Kiro capabilities including multi-step coding tasks, codebase- and team-standards context, review checkpoints before changes land, and correctness checking via property-based testing. It attributes the efficiency gain to Kiro grounding the model in requirements and task context up front, so it "arrives at working solutions faster, with fewer missteps."
No pricing, rollout timeline, regional availability, or per-variant guidance is given, and no third-party benchmark results accompany the Terminal-Bench figure. The GPT‑5.6 launch itself was announced separately; this is the downstream integration. OpenAI and AWS said they will continue tuning OpenAI models in Kiro.