Quick answer: GPT-5.6 is now available in Kiro in three versions. Choose Terra for most everyday coding, Luna for high-volume low-cost tasks, and Sol for difficult refactors or long-running terminal work. All three have a 272K context window in Kiro, but their current credit multipliers differ sharply.
GPT-5.6 Kiro model picker
| Choose | Best for | Kiro credit multiplier |
|---|---|---|
| Terra | Routine multi-step development and a sensible daily default | 1.0x |
| Luna | Repeated fixes, boilerplate and throughput-first work | 0.1x |
| Sol | Hard refactors, complex terminal tasks and sustained planning | 2.4x |
| Auto | When you want Kiro to balance cost and quality for you | 1.0x baseline |
Practical default: start with Terra. Drop to Luna for straightforward volume work, or move to Sol only when the task genuinely needs deeper planning and tool coordination.
OpenAI announced the Kiro integration on August 24, 2026. It brings the Sol, Terra and Luna tiers into Kiro workflows for planning, building, reviewing and testing software. OpenAI also says Kiro’s spec-driven structure gives the models requirements, technical designs and task context before implementation begins. Read OpenAI’s announcement.
How to use GPT-5.6 in Kiro
- Open Kiro in the IDE, CLI or another supported Kiro surface.
- Open the model selector in the chat interface.
- Select GPT-5.6 Terra, Luna or Sol. Your choice applies to later messages in that conversation.
- Choose reasoning effort if needed. In the IDE, use the model selector’s Effort panel. In the CLI, use
/effortor launch with the--effortflag. - Give Kiro a testable task. State the desired result, constraints, files that may change and the command that proves completion.
- Review the diff and run checks before merging generated changes.
Kiro’s current documentation says model switching is available from the chat dropdown, while reasoning-effort controls are available in the IDE and CLI. Check the live Kiro model documentation because availability and multipliers can change.
Sol vs Terra vs Luna: which should you choose?
Choose GPT-5.6 Terra for everyday agentic coding
Terra is the balanced tier. Use it for feature work, bug fixes that span several files, test creation, routine migrations and implementation from an existing specification. Its current 1.0x multiplier matches Kiro’s Auto baseline, making it the easiest manual starting point when you specifically want GPT-5.6.
Choose GPT-5.6 Luna for volume and credit efficiency
Luna has the lowest current GPT-5.6 multiplier in Kiro at 0.1x. It is the practical option for repetitive edits, simple test generation, documentation updates, code explanations and other high-frequency work where maximum reasoning depth is unnecessary.
Do not choose a cheaper model blindly for risky changes. Authentication, payments, permissions, data migrations and security-sensitive code still need careful human review and complete tests.
Choose GPT-5.6 Sol for the hardest work
Sol is Kiro’s recommended GPT-5.6 tier for long-horizon refactors and complex terminal work requiring sustained planning and tool coordination. Its current 2.4x multiplier makes it substantially more credit-intensive than Terra or Luna, so reserve it for tasks where retries and incomplete reasoning would cost more than the stronger model.
A copy-paste prompt for Kiro
Use this template to give any of the three models a bounded, verifiable task:
You are working in this repository to complete one defined change.
Goal:
[Describe the user-visible result.]
Constraints:
- Preserve existing public behavior unless explicitly changed.
- Do not modify files outside [allowed paths].
- Follow the repository's existing style and architecture.
- Do not expose secrets or weaken validation.
Before editing:
1. Inspect the relevant implementation and tests.
2. Summarize the smallest safe approach.
3. Identify any assumption that could change the solution.
Then:
1. Implement the change.
2. Add or update tests for success and failure paths.
3. Run [exact test/lint/build commands].
4. Report changed files, command results and any remaining risk.
Definition of done:
[Add measurable acceptance criteria.]Credit and pricing notes
Kiro measures work in credits rather than presenting these multipliers as direct per-token prices. The current paid plans listed by Kiro are Pro at $20 per user/month for 1,000 credits, Pro+ at $40 for 2,000 credits, Pro Max at $100 for 5,000 credits and Power at $200 for 10,000 credits, before applicable taxes. Paid users can buy add-on credits at $0.04 each. Verify current Kiro pricing before purchasing.
A multiplier is relative usage, not a promise that every prompt consumes a fixed number of credits. Kiro says simple prompts may use less than one credit, while complex work such as executing a specification can use more.
Important region and data-processing note
Kiro’s model documentation currently says GPT-5.6 requests are served from the United States regardless of the Kiro profile region. Teams with residency, regulatory or contractual requirements should review Kiro’s data-protection documentation before selecting these models. Model availability can also vary by country or region.
What changed for Kiro users?
- OpenAI models are now available directly in Kiro through the GPT-5.6 family.
- Users can choose among a capability-first, balanced or efficiency-first GPT-5.6 tier.
- The three tiers share a 272K context window in Kiro.
- Kiro can apply the models to spec-driven work grounded in requirements, designs and executable tasks.
OpenAI reports that its testing on Terminal-Bench 2.1 found GPT-5.6 Terra completed successful tasks in Kiro at roughly an 82% cost reduction. Treat that as a vendor-reported test result, not a guarantee for a particular repository or workload.
Frequently asked questions
Is GPT-5.6 available in Kiro now?
Yes. OpenAI says the GPT-5.6 family is available in Kiro, including Sol, Terra and Luna. Actual model availability can vary by country, region and plan.
What is the best GPT-5.6 model for Kiro?
Terra is the best starting point for routine multi-step coding. Use Luna when cost and throughput matter most, and Sol for the hardest long-running tasks.
Does GPT-5.6 in Kiro have a free tier?
Kiro’s pricing page describes GPT-5.6 as a premium model available to paid-plan users. Check the model selector and current pricing page for your account because access can vary by region and may change.
How much context do the Kiro GPT-5.6 models support?
Kiro’s current model table lists a 272K context window for Sol, Terra and Luna.
Should I use Kiro Auto or choose Terra manually?
Use Auto if you want Kiro to route tasks across models while balancing quality and cost. Choose Terra manually when you specifically want the balanced GPT-5.6 tier for the conversation.