Available models

Every model in Kiro is described below with its strengths, ideal use cases, and lifecycle status. For model comparisons, selection guidance, inference scope, and endpoint locations, see the Models overview.

Choose Sol for your hardest multi-step work, especially long-horizon refactors and complex terminal tasks. It is OpenAI's flagship GPT-5.6 tier for frontier reasoning, coding, long-horizon planning, and workflows that require iteration and tool coordination. Sol scores 80 on the Coding Agent Index and 88.8% on Terminal-Bench 2.1, above Claude Fable 5 on both while using less than half the output tokens and taking less than half the time. Sol has a 4.4x Kiro credit multiplier for requests up to 272K tokens.

Choose Terra for routine multi-step development when you want a middle ground between Sol's maximum capability and Luna's speed and Kiro credit efficiency. Terra scores 77.4 on the Coding Agent Index, compared with Claude Fable 5's 77.2. OpenAI separately describes Terra as GPT-5.5-competitive at half the API cost. That API comparison does not determine Kiro credit use; Terra has a 2.2x Kiro credit multiplier for requests up to 272K tokens.

Choose Luna for high-frequency agentic tasks where speed, throughput, and Kiro credit efficiency matter more than maximum capability. OpenAI describes Luna as its fastest and lowest-cost GPT-5.6 API tier. Luna scores 74.6 on the Coding Agent Index, compared with Claude Opus 4.8's 72.5. Luna has a 0.6x Kiro credit multiplier for requests up to 272K tokens, the lowest of the three GPT-5.6 tiers.

All three GPT-5.6 tiers have a 1M token context window in the IDE, CLI, and Web. You can include longer codebases, documents, and multi-turn agent histories in a request.

GPT-5.6 Sol, Terra, and Luna are generally available with full Kiro support across the Kiro IDE, CLI, Crew, and Web for Pro, Pro+, Pro Max, and Power customers in the US and EU geographies. Expanded capacity supports more dependable access for production workloads. GPT-5.6 Sol now has Active lifecycle status.

GPT-5.6 pricing follows the two-tier context model used by OpenAI and Amazon Bedrock: requests up to 272K tokens are billed at the short context rate listed above, and requests over 272K tokens are billed at double that rate (Sol 8.8x, Terra 4.4x, Luna 1.2x). Learn more.

Anthropic's most capable agentic coding model for long-running, context-heavy work. Fable 5.1 has moved beyond Preview and is generally available to Kiro Enterprise organizations with full Kiro support. Fable 5.1 traces problems to root causes, maintains context and intent across complex tasks, and improves reliability for multi-file refactors, long autonomous runs, and security or systems work. It has a 1M context window and a 6x credit multiplier. Inference runs only in US East (N. Virginia) (us-east-1).

Fable 5.1 can help discover software vulnerabilities with safeguards tuned to reduce interruptions during legitimate defensive security work. Requests for exploit generation, binary-based vulnerability scanning, and penetration testing continue to use Opus-class models in Kiro.

Enterprise users cannot see Fable 5.1 until an administrator enables model access management, selects Fable 5.1 in Manage approved list, and saves the list. Users must restart the IDE or CLI, or refresh Kiro Web, before it appears in the model selector.

All Fable 5.1 traffic is retained for up to 30 days (unless otherwise required by law) for automated offline abuse detection. Classifier-flagged traffic will be subject to potential human review performed by AWS, and we may review flagged traffic for abuse. See Data protection.

Kiro's model router. Auto routes your requests to frontier models with automatic fallback capabilities, so you get reliable top-tier quality without having to pick a model yourself. Free tier users get Claude Sonnet-class quality or better, while paid tiers get Claude Opus-class quality or better. Auto maintains a high quality bar to ensure results compare to or exceed the individual models available to you.

Auto is treated as its own selectable option under model governance: administrators can allow or block Auto for their users. When Auto is enabled, it may route requests to any Active model available to the user's Kiro profile, including models outside the administrator's approved list. Auto has the Geography inference scope shown in the Inference endpoint regions table and does not route to Experimental models. To limit processing to models on the approved list, block Auto and set an approved model as the default.

Anthropic's latest Opus model for demanding agentic coding and knowledge work. Anthropic reports stronger performance on long, complex coding tasks, self-verification described by early testers, and clearer communication that puts the most important information first. It performs at the level of Claude Fable 5.1 on most work, according to Anthropic. Claude Opus 5.5 is generally available with full Kiro support and Active lifecycle status across the Kiro IDE, CLI, Crew, and Web for Pro, Pro+, Pro Max, and Power customers in the US and EU geographies. The model has a 1M-token context window and a 2.0x Kiro credit multiplier. Learn more.

Anthropic's strongest agentic coding model. Sets new state-of-the-art results on leading coding benchmarks, more than doubling Opus 4.8's performance while approaching Fable 5 at half the cost. Excels at multi-file features, large refactors, and end-to-end feature work - completing full tasks rather than leaving stubs or placeholders. Opus 5 verifies its own work, iterates carefully until it succeeds, and introduces meaningful gains in multi-agent coordination with fewer conflicts between parallel agents. Ships with 1M context window and a 2.2x credit multiplier. Learn more.

Anthropic's most honest Opus model. Around 4x less likely than its predecessor to let flaws in generated code pass unremarked - it flags uncertainties and pushes back when evidence is thin rather than confidently claiming progress. Tool calling is meaningfully more efficient, using fewer steps for the same intelligence. Early testers report stronger judgment on when to act vs. when to ask, and better follow-through on long-running agentic tasks. Ships with 1M context window and 128K max output. Learn more.

Introduces adaptive thinking: the model automatically scales its internal reasoning based on task complexity. Simple questions get fast responses; complex architectural problems get deeper analysis, without you configuring anything. Beyond that, Opus 4.7 follows instructions more precisely, verifies its own outputs before reporting back, and supports 3x higher resolution vision for working with dense screenshots and diagrams. Learn more.

Top scores on Terminal-Bench 2.0 and SWE-bench Verified for agentic coding. Stays productive over longer sessions without context drift and handles multi-million-line codebases, planning upfront and adapting as needed. Strong debugging and code review capabilities let it catch its own mistakes through careful planning and self-correction. Learn more.

Handles tradeoffs and ambiguity well when working across multiple systems. Strong single-shot accuracy on complex problems where you need a correct answer on the first attempt, without iterative back-and-forth. Well suited for sophisticated software development challenges that span service boundaries. Learn more.

Anthropic's fastest Sonnet model to date and a clear upgrade over Sonnet 5. Anthropic reports more than 30% faster output, performance close to Opus 5.5 on real-world knowledge work, and clearer writing that early testers describe as a better partner for collaboration. Well suited for scoped coding, bug fixes, code review, agents and subagents, and scoped work on documentation, slides, and spreadsheets. Claude Sonnet 5.5 is generally available across the Kiro IDE, CLI, Crew, and Web for Pro, Pro+, Pro Max, and Power customers. The model has a 1M-token context window, adaptive thinking on by default, and a 1.3x Kiro credit multiplier. Cybersecurity safeguards are similar to Opus 5.5: routine development work is unaffected, while some narrow, high-risk requests may be refused.

Anthropic's most agentic Sonnet model before Sonnet 5.5. Performance approaches Opus 4.8 on reasoning, tool use, coding, and knowledge work, at Sonnet-class pricing. Plans before it edits, runs longer without supervision, and checks its own output without being asked. Early testers consistently describe it as finishing complex tasks where prior Sonnet models would stop short. Well suited for spec-driven workflows where you need higher-fidelity implementation across broad changes, and for multi-step agentic tasks in the CLI and Web that run to completion with less supervision. Learn more.

A full upgrade from Sonnet 4.5 that approaches Opus 4.6 intelligence while being more token efficient. Excels at iterative development workflows and maintains context across long sessions. Handles both lead agent and subagent roles in multi-model pipelines, making it well-suited for teams using Kiro powers and custom subagents. Learn more.

Strong agentic coding with extended autonomous operation - can work independently for hours with effective tool usage. Improved planning, system design, and security engineering compared to Sonnet 4.0. Learn more.

Direct access to Anthropic's Claude Sonnet 4.0 for users who prefer consistent model selection. Same model for all interactions with no routing or optimization layers. Full control and complete transparency, with predictable behavior for workflows that depend on specific model characteristics. Claude Sonnet 4.0 reaches end of life on October 14, 2026. Learn more.

Choose Haiku 5.5 for high-volume work requiring quick responses. It has a 1M-token context window. Claude Haiku 5.5 is available with experimental support across the Kiro IDE, CLI, Crew, and Web for Pro, Pro+, Pro Max, and Power customers in the US and EU geographies. Requests up to 100K tokens have a 0.06x Kiro credit multiplier. Requests that exceed 100K tokens are billed at a 0.3x multiplier.

Haiku 5.5 uses cross-region inference within the selected geography: US-profile requests stay within the US, and EU-profile requests stay within the EU.

Anthropic's fastest model with near-frontier performance. Matches Sonnet 4 performance across reasoning and coding at more than twice the speed. Near-frontier intelligence at one-third the cost, and the first Haiku model with extended thinking capabilities. Learn more.

Moonshot AI's open-weight model for long-running coding and knowledge work. Kimi K3 accepts text and images and has a 1M-token context window. It has a 2.4x credit multiplier and experimental support. Kimi K3 is available to customers using Kiro's US geography. It uses cross-region inference within the US geography, so requests stay in supported US AWS Regions. See the Amazon Bedrock model card.

Kimi K3 appears in the model picker for Pro, Pro+, Pro Max, and Power subscribers using Kiro's US geography. It is not available on the Free tier.

Enterprise users cannot see Kimi K3 until an administrator enables model access management, selects Kimi K3 in Manage approved list, and saves the list. Users must restart the IDE or CLI, run kirocrew restart for Crew, or refresh Kiro Web before Kimi K3 appears in the model picker.

Open weight model that matches frontier-class coding performance at a fraction of the cost. Trained with reinforcement learning across hundreds of thousands of real-world environments, delivering strong results across the full development lifecycle from system design to code review. It has a 0.25x credit multiplier. Learn more.

Open weight sparse mixture-of-experts model with a 200K context window, designed for complex systems engineering and long-horizon agentic tasks. Excels at processing repository-scale context and maintaining coherence during multi-step tool use across large codebases. Well suited for cross-file migrations, full-stack feature development, and legacy refactoring where the model needs to hold the full picture. It has a 0.5x credit multiplier. Learn more.

Open weight model best suited for agentic workflows and code generation. Handles long tool-calling chains, stateful sessions, and multi-step reasoning well. It has a 0.25x credit multiplier. Learn more.

Open weight model best suited for multilingual programming and UI generation. Delivers strong results across Rust, Go, C++, Kotlin, TypeScript, and others. It has a 0.15x credit multiplier. Learn more.

Open weight model purpose-built for coding agents with 256K context and strong error recovery. Works especially well for long agentic coding sessions. It has a 0.05x credit multiplier. Learn more.

Each model family brings a distinct working style. Understanding these differences helps you pick the right one for the task.

GPT-5.6 models can write lightweight programs that coordinate tools and process intermediate results instead of making separate round-trips for each step. This reduces token use and dead ends, especially for spec implementation, terminal workflows, and multi-file refactors. All three tiers support longer unsupervised execution and use hidden chain-of-thought, so you see the final output rather than internal reasoning steps. Choose Sol for the hardest long-horizon work, Terra for balanced routine development, and Luna when speed and Kiro credit efficiency matter most.

Opus models plan thoroughly before acting, considering multi-step approaches and edge cases upfront. They catch their own mistakes through careful review and flag uncertainty rather than confidently claiming progress. Starting with Opus 4.7, reasoning depth scales automatically based on task complexity - simpler prompts get fast responses while harder problems receive deeper analysis without you configuring anything. Opus 5.5 strengthens long-horizon coding, self-verification, and clear communication while lowering the Kiro credit multiplier from Opus 5's 2.2x to 2.0x. Opus excels at sustained multi-file work, complex debugging, and architecture decisions where self-verification matters.

Sonnet models balance strong agentic performance with lower Kiro credit use than Opus models. Sonnet 5.5 is a clear upgrade over Sonnet 5 and produces output more than 30% faster, making it a strong choice for scoped coding, bug fixing, review, and agent or subagent work. Earlier Sonnet models are more conservative and stick closer to what you asked for, making them predictable choices for focused, well-scoped work.

Open weight models span credit multipliers from 0.05x to 2.4x. Kimi K3 handles long-running coding and knowledge work with a 1M context window and image input. MiniMax M2.5 delivers near-Opus results across the development lifecycle. GLM-5 handles repository-scale context for cross-file migrations and full-stack work. DeepSeek 3.2 handles long tool-calling chains and multi-step reasoning. Qwen3 Coder Next is the most cost-effective option with strong error recovery for extended CLI sessions.

Models in Kiro move through stages. Each stage reflects the model's maturity and the level of support you can expect.

Kiro generally schedules model EOL alongside Amazon Bedrock. Kiro may set a different date. When Kiro schedules removal, the picker shows [EOL] and its date. If your work depends on a model marked [EOL], choose another available model before continuing. After Kiro removes a model, requests to it fail. Kiro does not switch your selected model automatically. See the Quick comparison table for all models and plan access.

Kiro publishes known end-of-life dates here when announced.

The following models are "Kiro Excluded Models":

• Moonshot Kimi K3
• Models designated as "experimental", "pre-release", "preview", or "beta"

When you elect to use a Kiro Excluded Model, the output indemnity provided in Section 50.10 of the AWS Service Terms does not apply.

The model selector and the model lifecycle table show which models are experimental. Enterprise administrators can control which models their users can choose with model governance.