Overview
Anthropic changed the default effort level for Opus 4.6 in Claude Code to 'medium' for Max and Team subscribers, effective with version 2.1.68. The change appears in the changelog entry for that release and applies specifically to interactive sessions under those subscription tiers.
Prior to this change, the default effort for API-key, Bedrock/Vertex/Foundry, Team, and Enterprise users had been set to 'high' as of version 2.1.94 — so this represents a tier-specific rollback for Max and Team users on Opus 4.6. Effort level controls how much computation the model allocates per response, affecting reasoning depth, response length, and how thoroughly Claude works through multi-step problems. This is a model behavior change, not a pricing or policy update.
What this means for brands
If your team or agency uses Claude Code on Max or Team plans to generate content, research, or structured brand outputs, responses from Opus 4.6 will now be shallower by default. Queries that previously returned detailed competitive breakdowns, multi-angle brand analyses, or thorough citation-style summaries may now return shorter or less comprehensive outputs without any visible signal that the effort level dropped. This is not a model swap, so diagnostics based on model name alone will miss it.
For teams using Claude Code in automated pipelines — brand monitoring prompts, content audits, query-testing workflows — the quality floor just shifted. Outputs that passed quality checks under the prior default may fall below threshold now. Any benchmark you ran against Opus 4.6 before this change should be treated as stale.
What to do
Re-run your standard brand query set against Opus 4.6 this week on a Max or Team account and compare output depth to your prior baseline. If you need the previous behavior, run '/effort high' at the start of a Claude Code session to override the new default — this setting persists for the session duration. For automated pipelines, add 'effort: high' to any skill or slash command frontmatter that drives brand research or content tasks. If you manage enterprise or API-key deployments, this change does not affect you, but verify your tier before assuming outputs are unchanged.