Claude Code users can swap between AI models on the fly with a simple slash command, a flag, or by tweaking shell variables. The quick /model command updates the active model instantly without restarting the terminal. This ease of switching helps developers test different model behaviors during coding sessions.

To change the model for a single session, launch Claude Code with the --model flag, specifying either the base model or the newer Opus variant. For a permanent default, users edit their shell profile—adding a line like export CLAUDE_MODEL=opus—and then restart the terminal. This makes the chosen model the fallback for all future runs.

These configuration options give developers granular control over which Claude model powers their IDE. The ability to toggle between the standard and Opus models without downtime streamlines experimentation and can reduce time spent debugging prompt‑model mismatches. By locking a model into the shell, teams ensure consistency across projects and CI pipelines.

Because Opus is only available when users enable additional usage tiers, the documentation clarifies that enabling the flag in the shell profile does not automatically unlock the model; customers must subscribe to the higher‑tier plan first. Once activated, the model behaves like any other, providing larger context windows and faster inference for complex code generation tasks.