Key takeaways:
- GPT-5.6 is now the default model powering Kiro, OpenAI's agentic IDE for developers.
- It supports the four core coding-agent tasks Kiro is built around: plan, build, review, and test.
- The upgrade is pitched primarily as a price-performance improvement, not a headline capability jump.
- Existing Kiro users get the new model inside current workflows, with no migration or setup changes required.
Kiro users got a quiet upgrade this week. OpenAI has rolled out GPT-5.6 inside Kiro, the agentic IDE, and it's now the model doing the heavy lifting when you ask it to plan a feature, write the code, review a pull request, or draft tests. The pitch is straightforward: better price-performance. Same categories of task, less spend per meaningful output, according to OpenAI's own note.
If you already work in Kiro, you don't need to switch anything; the newer model shows up in existing workflows. If you don't, this is one more reason to take a look. Below is what OpenAI has actually said, what price-performance means in this context, and how GPT-5.6 sits next to the other coding-focused models developers reach for day to day.
What OpenAI actually announced
The announcement is short. According to OpenAI, GPT-5.6 is now available in Kiro, helping developers plan, build, review, and test software with better price-performance. That's the entire scope of the update as published: a model swap inside a specific product, aimed at a specific workflow.
No new benchmarks. No separate API tier. No new feature surface in Kiro itself has been shared alongside the note. If you're waiting for eval numbers or a detailed comparison to GPT-5.5, they aren't in this release post.
Where GPT-5.6 fits inside Kiro's workflow
Kiro is structured around four agent-driven stages, and GPT-5.6 is the model behind each of them now.
- Plan. Turning a vague ticket or spec into a concrete task breakdown the agent can execute.
- Build. Writing the actual code changes across files.
- Review. Reading a diff or PR and flagging issues before a human does.
- Test. Generating and running tests to catch regressions.
That end-to-end shape is what separates an agentic IDE from a plain autocomplete assistant, and it's also why the price side of price-performance matters so much. An agent that plans, edits, reviews, and tests burns tokens at every step. Small per-call savings compound quickly.
Why price-performance is the headline, not raw capability
Coding agents in 2026 are converging on similar top-line capability. What's still moving is cost per useful task. That's the axis OpenAI is highlighting here, and it's the same axis pushing releases like and into the market.
For teams evaluating whether to standardize on an agentic IDE at all, the math is usually about token spend across a real engineering week, not a single benchmark score. If GPT-5.6 shifts that math even modestly inside Kiro, it matters more than a two-point bump on a coding eval. Cost is often the deciding factor between building on a hosted agent and staying with in-house tooling, a tradeoff we've covered in.
What Kiro users should do next
Nothing, in the short term. The model change is live inside the product, so your existing planning, build, review, and test flows are already routing through GPT-5.6. Worth watching over the next few sprints: whether your monthly usage bill drops for the same volume of agent work, and whether review-stage suggestions feel tighter. Those are the two places a price-performance upgrade tends to show up first in day-to-day use.
Common questions
Is GPT-5.6 available outside Kiro?
The announcement covers availability inside Kiro specifically. OpenAI hasn't detailed broader API or ChatGPT rollout in this post, so treat Kiro as the confirmed surface for now.
Do I need to change my Kiro setup to use it?
No. The model is enabled inside existing Kiro workflows. There's no separate toggle, plan change, or migration step called out in the release.
Does this replace GPT-5.5 in Kiro?
OpenAI's note frames GPT-5.6 as the model now powering Kiro's plan, build, review, and test tasks. It doesn't spell out the exact deprecation path for prior versions, so watch Kiro's own release notes for that detail.
Related service: AI Agency