Model selection

Kiro provides multiple powerful AI agent options to handle your development tasks: Auto, Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.5, Claude Sonnet 4.0, and Claude Haiku 4.5. Each offers distinct advantages depending on your needs and usage patterns.

Available models

Auto (recommended)

Auto is Kiro's default intelligent model router that combines multiple frontier models with advanced optimization techniques.

Key benefits:

• Cost-effective – Approximately 23% less expensive than direct Sonnet 4 usage
• Smart routing – Automatically chooses the optimal model for each task
• Consistent quality – Delivers Sonnet 4-level results across different task types
• Plan efficiency – Makes your usage limits go further

What model does Auto use?

Auto uses best in class LLM models (Claude Sonnet 4 and alike) to provide you the best quality for the type of tasks assigned to the agent. We maintain a very high bar to ensure that the quality of what is offered under Auto compares to or exceeds the quality of separate models made available to our users.

Claude Opus 4.6 (experimental)

Anthropic's smartest model with state-of-the-art coding and agentic capabilities. Opus 4.6 plans more carefully, sustains tasks for longer, operates reliably in larger codebases, and catches its own mistakes through better code review and debugging.

Key benefits:

• State-of-the-art coding – Top scores on Terminal-Bench 2.0 and SWE-bench Verified for agentic coding
• Long-horizon tasks – Stays productive over longer sessions without context drift
• Large codebase navigation – Handles multi-million-line codebases like a senior engineer, planning upfront and adapting strategy
• Better self-correction – Improved debugging and code review skills to catch its own mistakes
• Deeper reasoning – Thinks more carefully on complex problems, revisiting reasoning before settling on answers

Claude Opus 4.5

Anthropic's most intelligent model combining maximum capability with practical performance. Ideal for complex specialized tasks, professional software engineering, and advanced agents.

Key benefits:

• Maximum intelligence – Step-change improvements in reasoning, coding, and problem-solving
• Practical performance – More accessible price point than previous Opus models
• Complex reasoning – Better balance of tradeoffs and ambiguity across multiple systems
• Production-ready – Handles the most sophisticated software development challenges

Claude Sonnet 4.5

Anthropic's best model for complex agents and coding, with the highest intelligence across most tasks.

Key benefits:

• Coding excellence – Advanced state-of-the-art on SWE-bench Verified
• Agent capabilities – Extended autonomous operation for hours with effective tool usage
• Enhanced reasoning – Improved planning, system design, and security engineering

Claude Sonnet 4.0

Direct access to Anthropic's Claude Sonnet 4.0 model for users who prefer consistent model selection or have specific requirements for using this particular model.

Key benefits:

• Predictable behavior – Same model for all interactions
• Direct access – No routing or optimization layers
• Full control – Complete transparency in model selection

Claude Haiku 4.5

Anthropic's fastest and most intelligent Haiku model with near-frontier performance. Delivers intelligence matching Sonnet 4 at significantly lower cost and faster speed.

Key benefits:

• Near-frontier intelligence – Matches Sonnet 4 performance across reasoning and coding
• Blazing speed – More than twice the speed of Sonnet 4
• Cost-effective – Near-frontier intelligence at one-third the cost
• Extended thinking – First Haiku model with advanced reasoning capabilities

Cost comparison

Understanding the credit consumption differences:

Model	Credit Usage	Example Task Cost
Claude Opus 4.6	2.2x	22 credits
Claude Opus 4.5	2.2x	22 credits
Claude Sonnet 4.5	1.3x	13 credits
Claude Sonnet 4.0	1.3x	13 credits
Auto	1.0x	10 credits
Claude Haiku 4.5	0.4x	4 credits

Choosing the right model

Opus 4.6 (experimental)

Consider using Opus 4.6 when:

• Spec-driven development – Working with Kiro specs for structured, multi-step implementations
• Long-running agentic tasks – Tasks requiring sustained focus and multi-step planning
• Large codebase work – Navigating and modifying code across many files and systems
• Complex debugging – Problems requiring careful exploration and self-correction

Opus 4.5

Consider using Opus 4.5 when:

• Maximum intelligence needed – Most complex specialized tasks requiring top-tier reasoning
• Professional software engineering – Sophisticated production development challenges
• Advanced agents – Long-running autonomous tasks with complex decision-making
• Critical implementations – High-stakes features where quality is paramount

Sonnet 4.0 & Sonnet 4.5

Consider using Sonnet 4.0 or Sonnet 4.5 when:

• Consistency is critical – You need predictable model behavior
• Specific requirements – Your workflow depends on Sonnet 4's particular capabilities
• Model transparency – You prefer knowing exactly which model handles each request
• Budget flexibility – Higher costs aren't a primary concern

Auto

Consider using Auto when:

• Cost efficiency matters – You want to maximize your plan's value
• General development work – Most coding, debugging, and planning tasks
• Variable task types – Working on diverse projects with different requirements
• Plan optimization – You want your limits to stretch further

Haiku 4.5

Consider using Haiku 4.5 when:

• Speed is critical – You need fast responses for interactive experiences
• Cost efficiency matters – Near-frontier intelligence at one-third the cost
• High-volume processing – Cost-effective intelligence for large-scale deployments
• Real-time applications – Fast turnaround for user-facing features

How to switch models

In the chat interface

1. Access model selector

   • Look for the model dropdown in the chat interface
   • Default selection is Auto

2. Choose your model

   • Select Claude Opus 4.6 for large-scale codebases and spec-driven workflows (experimental)
   • Select Claude Opus 4.5 for complex production tasks
   • Select Claude Sonnet 4.5 for the latest Anthropic model
   • Select Claude Sonnet 4.0 for direct access
   • Select Auto for intelligent routing
   • Select Haiku 4.5 for fast, cost-effective tasks

3. Confirmation

   • Your selection applies to subsequent messages
   • Current conversation context is maintained

Best practices

Maximizing efficiency

• Start with Auto – Use it as your default for most tasks
• Use Haiku for speed – Near-frontier intelligence with fast responses and lower cost
• Use Sonnet for agents – Best for complex coding and autonomous workflows
• Reserve Opus for maximum intelligence – Use for the most complex specialized tasks
• Monitor usage – Track how different models affect your plan consumption
• Experiment – Try different models for similar tasks to compare results

Cost management

• Plan accordingly – Factor model choice into your tier selection
• Track patterns – Understand which tasks benefit most from each model
• Optimize workflows – Adjust development practices based on model strengths
• Consider overages – Enable if you need flexibility beyond plan limits