What exactly changes when you use /model opusplan in Claude Code?

Answers 1

첫째StaffLv.67/20/2026
Hey, mod team here. You got it exactly right. /model opusplan does run Opus during the planning stage and then switches to Sonnet once you move into the execution phase—that's the hybrid mode. Here's the difference from /model opus: 1. /model opus keeps everything on Opus the whole conversation—planning, executing, all of it. 2. /model opusplan uses Opus just for the planning stage when you need the heavy reasoning power, then drops to Sonnet for the actual implementation. The conversation context stays intact across the model switch, so Sonnet still has access to everything Opus planned and can execute from there. About token consumption: Opus is a bigger model that burns through tokens faster than Sonnet—that's publicly known. Since opusplan only uses Opus for the planning phase instead of the whole conversation, you do burn fewer tokens overall compared to running everything on Opus. But there's no public number on exactly how much more opusplan uses versus pure Sonnet. The bigger your planning phase is—like if you're doing a complex architecture design or large refactor—the more tokens you'll burn, since that's where Opus comes in. One note: if your org has a policy that restricts Opus access, opusplan's planning stage will use the latest Opus version your org does allow. If all Opus versions are blocked, it falls back to Sonnet. Mileage varies depending on what kind of work you're doing, so if you try it out and want to share your experience in the comments, that'd be super helpful. For a breakdown of how token consumption works and ways to optimize it, check out the guide below. https://usingclaude.com/ko/pricing/plans/claude-usage-limits-guide
Report