Use /model opusplan

/model opusplan uses Opus for planning and Sonnet for implementation.

/model opusplan

Why it is useful

• Opus is stronger for planning, architecture, debugging strategy, and complex reasoning.
• Sonnet is usually faster and more cost-effective for implementation.
• The workflow gives you a stronger plan first, then efficient execution.

Typical workflow

1. Run /model opusplan.
2. Enter Plan Mode.
3. Let Opus create the plan.
4. Approve the plan.
5. Claude Code uses Sonnet for implementation.

The cache trade-off

Every time you enter or leave plan mode, opusplan switches model, and the next request re-reads the conversation uncached. It works best when you plan once at the start, not when you hop in and out of plan mode. See Prompt cache pitfalls.

1M context

On Max, Team, and Enterprise plans, Opus is automatically upgraded to 1M context, and opusplan gets that upgrade in plan mode. Sonnet 5 always runs with the 1M window on the Anthropic API. So you rarely need a suffix.

If you are not on an auto-upgrade tier, /model opusplan[1m] forces 1M context for both phases. Setting it with /model requires v2.1.265 or later. On earlier versions, use the --model flag or the model setting.