Models & Reasoning Effort

Different tasks call for different models and different “thinking depths.” This chapter helps you pick the right model and set it as your default.

First, see what models are available

# Open the interactive menu inside a conversation
/model

# Or list them from the command line
kiro-cli chat --list-models
kiro-cli chat --list-models --format json-pretty

Switch models

# Switch directly in a conversation (supports partial matching, case-insensitive)
/model claude-sonnet-4

# Specify at launch
kiro-cli chat --model claude-opus-4.8 "Explain this codebase"

Set a default model

Want every new session to use the same model? Set it like this:

kiro-cli settings chat.defaultModel claude-opus-4.8

You can also save the model your current session is using as the default:

/model set-current-as-default

When it comes to picking a model, the priority order is (highest to lowest):

1. --model flag
2. The model restored from a --resume session
3. The chat.defaultModel setting
4. System default

Reasoning effort

effort controls how “deeply” the model thinks. The higher you go, the more carefully it thinks — but it’s also slower and burns more usage:

# Adjust in a conversation
/effort high

# Specify at launch
kiro-cli chat --effort max "Do a security review of this PR for me"

Remember a default effort for each model

In ~/.kiro/settings/cli.json, use chat.modelDefaults to set a default effort for a specific model. For the Claude family, write it under output_config.effort:

{
  "chat.defaultModel": "claude-opus-4.8",
  "chat.modelDefaults": {
    "claude-opus-4.8": { "output_config": { "effort": "max" } }
  }
}

That way every new session applies it automatically, no need to pass the flag every time.

The effort priority order (highest to lowest): --effort flag → the loaded session → the cli.json default → the model’s built-in default.

In the next chapter, we’ll save conversations so you can pick up where you left off later.