Launch Claude Code with a Specific Model Using --model

You already know you can switch models mid-conversation with /model — but you can also set it upfront when launching Claude Code from the command line.

# Use Opus for a complex architecture task
claude --model claude-opus-4-6 \
  "Design the data model for a multi-tenant SaaS billing system"

# Use Haiku for a quick, cheap lookup
claude --model claude-haiku-4-5-20251001 \
  "What does this regex match? /^[a-z0-9._%+-]+@[a-z0-9.-]+\.[a-z]{2,}$/"

The --model flag sets the starting model for your session. You can still switch with /model during the conversation if your needs change.

This is especially useful when you're scripting Claude Code invocations and want deterministic, reproducible behaviour — like in CI pipelines or automation scripts where you'd rather explicitly pin the model than rely on a default that might change.

# In a Makefile or automation script
review:
    git diff HEAD | claude --model claude-sonnet-4-6 -p \
      "Review this diff for bugs and security issues."

docs:
    claude --model claude-opus-4-6 \
      "Read the codebase and write comprehensive API documentation."

Matching model to task is a quick way to optimise for cost, speed, or quality depending on what the job actually needs.

Pick your model before you start — match the horsepower to the task.