Claude Code v2.1.280 and later can select Opus 5.5 through the opus alias or its full model name, and adjust the effort level with /effort, startup flags, or a settings file.
Suitable tasks: Selecting Opus 5.5 in a Claude Code session; configuring effort for tasks of different complexity, such as coding and planning; avoiding alias changes that could switch to another model as the provider or version changes.
Unsuitable tasks: Using the commands and settings below directly in Anthropic Messages API requests. API requests should use the model ID and output_config.effort specified in the API documentation.
Applicable model version: Claude Opus 5.5; Anthropic requires Claude Code v2.1.280 or later.
Applicable client, agent, or API: Claude Code CLI and its model settings. The provider affects alias mappings and model availability; on Microsoft Foundry, opus currently maps to Opus 4.6.
Recommended effort levels and parameters: Opus 5.5 defaults to medium. Evaluate low, medium, high, xhigh, and max for your tasks. The settings file's modelSettings supports low, medium, high, and xhigh, but does not accept max.
Select the Opus alias in a Claude Code session (entering this directly saves it as the default for subsequent sessions):
/model opusTo pin Opus 5.5 and avoid depending on future alias updates, specify the full model name:
claude --model claude-opus-5-5Set the effort level at startup:
claude --model claude-opus-5-5 --effort highSet the effort level for the current session, or clear the saved level and restore the model default:
/effort high
/effort autoYou can also run /effort to open the effort slider, or use the left and right arrow keys to adjust effort while selecting a model with /model.
Confirm that your Claude Code version is at least v2.1.280. On an older version, run claude update and restart.
Run /model to see the models currently available. Selecting opus uses the Opus model currently mapped by the provider. If you specifically need Opus 5.5, select the full model name claude-opus-5-5. Entering the command directly saves it as the default for subsequent sessions; to switch only the current session, press s in the model picker.
Run /effort, starting with the default medium; compare low, high, or other available levels on representative tasks.
To pin the model and effort when starting a session, use claude --model claude-opus-5-5 --effort high. The command-line --model and --effort flags apply only to the session started by that invocation.
Verify the model and effort level actually selected for the session, and record the Claude Code version, provider, task, effort level, and result.
The official documentation says Claude Code supports setting the model through the in-session /model <alias|name> command, the startup flag claude --model <alias|name>, the ANTHROPIC_MODEL environment variable, the model setting in a settings file, or ANTHROPIC_DEFAULT_MODEL.
The documentation lists the provider mappings for opus: Anthropic API, Claude Platform on AWS, Amazon Bedrock, and Google Cloud Agent Platform map to Opus 5.5; Microsoft Foundry maps to Opus 4.6. These mappings may change between versions.
Anthropic explicitly requires Claude Code v2.1.280 or later for Opus 5.5. The full model name example is claude-opus-5-5.
Opus 5.5 supports the low, medium, high, xhigh, and max effort levels, and defaults to medium. Lower levels are faster and less expensive; higher levels provide deeper reasoning for complex tasks. This is Anthropic's description, not an independent measurement by the collector.
CLAUDE_CODE_EFFORT_LEVEL can be set to an effort level name or auto. modelSettings can set an effort level for a model, but its accepted values are limited to low, medium, high, and xhigh; max is not valid for this setting key.
Opus 5.5 does not read the legacy top-level user-level effortLevel default. You can select an effort level for Opus 5.5 through /effort or /model. A top-level effortLevel in project, local, or managed settings, or supplied through --settings, can still apply to all models.
opus is a provider alias that can change over time. If a provider maps it to an older model, Anthropic recommends using the full model name or the corresponding ANTHROPIC_DEFAULT_OPUS_MODEL setting to pin a version. Do not assume that opus means Opus 5.5 with every provider.
Microsoft Foundry currently maps opus to Opus 4.6, so selecting opus there will not give you Opus 5.5. Whether the full model name is available also depends on whether the provider offers that model.
claude-opus-5-5 in these commands is the full model name accepted by Claude Code. For Anthropic Messages API settings, API model IDs, parameter structures, and effort configuration, follow the API documentation.
Effort affects reasoning depth, speed, and cost. The official page does not provide quality or cost measurements that can be generalized across task types; compare the levels on your own task set.
Record claude --version, the login method or provider, the full model name, effort level, task input, and completion result. Test /model opus and /model claude-opus-5-5 separately to confirm which model is actually resolved. Starting at medium, change the effort level on the same task set and compare accuracy, elapsed time, and usage. Alias resolution, model availability, and settings precedence may vary with the Claude Code version, provider, and organization policy; recheck the official page when reproducing the procedure.
Claude Opus 5.5