{"version":"2.1.317","anchor":"subagent-resume-keeps-its-picked-effort-model","canonical_anchor":"subagent-resume-keeps-its-picked-effort-model","heading":"Resumed subagents keep the effort level of the model they started with","tier":"notice","area":"Subagents","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.317\/e\/subagent-resume-keeps-its-picked-effort-model","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.317","markdown":"### Resumed subagents keep the effort level of the model they started with\n\nA resumed subagent now remembers which model its effort level was picked for and works out its effort against that model\n\n**What**\n\nA subagent is a helper that Claude Code starts to handle part of a task, and it can be resumed later. Effort is how much thinking the model is asked to do.\n\nResumed subagents now save and restore the model their effort level was picked for, and work out their effort against that model. Before, effort was worked out from whichever model the subagent ended up using. The rule for when a resume is treated as a reply also gained an extra condition.\n\nThis sits behind a check that the `CLAUDE_CODE_HARMONIC_RIDDLE` environment variable can also control.\n\n**Why**\n\nA resumed subagent should keep a consistent effort level even if the model was changed in between.\n\n- Area: Subagents\n- Tier: You'll notice\n- Useful: 2\/5\n- Signal: 1\/5\n- Scope: individual\n- Heads-up: no"}