LLM & Model

`model`

Type: string
Default: unspecified
Description: Override the default model to use for Claude Code. Examples: "claude-opus-4-6", "claude-sonnet-4-6", "claude-haiku-3-5".
Example:
```
{
  "model": "claude-sonnet-4-6"
}
```

`availableModels`

Type: string[]
Default: unspecified
Description: Restrict which models users can select via /model, --model, Config tool, or ANTHROPIC_MODEL. Does not affect the Default option.

Example:

{
  "availableModels": ["sonnet", "haiku"]
}

`modelOverrides`

Type: object
Default: unspecified
Description: Map Anthropic model IDs to provider-specific model IDs such as Bedrock inference profile ARNs. Each model picker entry uses its mapped value when calling the provider API.

Example:

{
  "modelOverrides": {
    "claude-opus-4-6": "arn:aws:bedrock:us-west-2:123456789012:inference-profile/anthropic.claude-opus-4-6-20250514-v1:0"
  }
}

`advisorModel`

Type: string
Default: unspecified
Description: Model consulted by the server-side advisor tool — a second-opinion model the main loop can invoke alongside the primary model. Managed interactively via /advisor <model> (set) or /advisor unset (clear). Has no effect unless the current main-loop model supports the advisor feature (see modelSupportsAdvisor). Configurability is further gated by canUserConfigureAdvisor.
Example:
```
{
  "advisorModel": "claude-opus-4-6"
}
```
Source: claude-code-main/utils/settings/types.ts:712-715; commands/advisor.ts.

`alwaysThinkingEnabled`

Type: boolean
Default: unspecified
Description: Enable extended thinking (Claude’s reasoning mode) by default for all sessions. When enabled, Claude will use extended thinking for tasks that benefit from deeper analysis.
Example:
```
{
  "alwaysThinkingEnabled": true
}
```

`effortLevel`

Type: string enum
Default: unspecified
Valid values (per public docs): "low", "medium", "high", "xhigh", "max". Available values depend on the model. "max" is session-only and not persisted (setting it via /effort max applies to this session only).
Valid values (in the claude-code-main source snapshot): "low", "medium", "high" (public builds) or "low", "medium", "high", "max" (ant-only builds). "xhigh" does not appear in this scrubbed source — same public-docs-vs-source pattern documented in commands/discrepancies-and-gaps. Trust the public docs for availability; the schema is guarded with .catch(undefined) so unknown values are silently ignored.
Description: Persisted effort level for supported models. Written by /effort <level> and cleared by /effort unset. /effort auto clears the setting to fall back to the default for the current model.
Overridden by: CLAUDE_CODE_EFFORT_LEVEL environment variable. When set, it takes effect for the current session and the settings value is not applied until the env var is unset.
Example:
```
{
  "effortLevel": "high"
}
```
Source: claude-code-main/utils/settings/types.ts:703-711; commands/effort/effort.tsx. Public reference: https://code.claude.com/docs/en/commands (entry for /effort).

`fastModePerSessionOptIn`

Type: boolean
Default: unspecified
Description: When true, fast mode does not persist across sessions. Each session starts with fast mode off, requiring users to enable it with /fast. The user’s fast mode preference is still saved.
Example:
```
{
  "fastModePerSessionOptIn": true
}
```

`disableAutoMode`

Type: string
Default: unspecified
Description: Hard-disable auto mode. Set to "disable" to prevent auto mode from being activated. Removes auto from the Shift+Tab cycle and rejects --permission-mode auto at startup.
Example:
```
{
  "disableAutoMode": "disable"
}
```

`disableBypassPermissionsMode`

Type: string
Default: unspecified
Description: Hard-disable bypass-permissions mode. Set to "disable" to prevent bypassPermissions mode from being activated. Disables the --dangerously-skip-permissions command-line flag. ⚠ Security

Example:

{
  "disableBypassPermissionsMode": "disable"
}

← Back to settings/README.md