Claude Opus 5.5 in Claude Code: What Changes for Your Workflow

<!– wp:paragraph –>
<p>Claude Opus 5.5 became the default Opus model in Claude Code within days of release. Here’s what actually changed, and what’s worth checking in your own setup.</p>
<!– /wp:paragraph –>

<!– wp:heading –>
<h2 class=”wp-block-heading”>What shipped</h2>
<!– /wp:heading –>

<!– wp:paragraph –>
<p>Anthropic released <strong>Claude Opus 5.5</strong> on 22 September 2026, the first model in the Claude 5.5 family, and it reached Claude Code the same week. Coverage of the launch says it performs close to Claude Fable 5.1 on most tasks while costing roughly 40% less to run than Opus 5. Sonnet 5.5 and Haiku 5.5 are expected to follow in the coming weeks.</p>
<!– /wp:paragraph –>

<!– wp:paragraph –>
<p>Claude Code 2.1.280 made Opus 5.5 the default Opus model, so picking <code>opus</code> in the model picker already gets you the new one. There’s nothing extra to install beyond updating Claude Code itself.</p>
<!– /wp:paragraph –>

<!– wp:heading –>
<h2 class=”wp-block-heading”>The numbers that matter for coding work</h2>
<!– /wp:heading –>

<!– wp:list –>
<ul class=”wp-block-list”><!– wp:list-item –>
<li><strong>1M-token context window</strong> – enough to hold a mid-sized repository, its tests, and a long conversation without constant compaction.</li>
<!– /wp:list-item –><!– wp:list-item –>
<li><strong>128k maximum output tokens</strong> – large refactors and generated files are less likely to get cut off part way through.</li>
<!– /wp:list-item –><!– wp:list-item –>
<li><strong>Always-on adaptive thinking</strong> – the model adjusts how much it reasons per step automatically, rather than you toggling a thinking mode.</li>
<!– /wp:list-item –><!– wp:list-item –>
<li><strong>Pricing</strong> – reported at $4 per million input tokens and $20 per million output tokens, against $5/$25 for Opus 5. Worth confirming against Anthropic’s own pricing page before you budget around it.</li>
<!– /wp:list-item –></ul>
<!– /wp:list –>

<!– wp:heading –>
<h2 class=”wp-block-heading”>Things worth checking after you update</h2>
<!– /wp:heading –>

<!– wp:list {“ordered”:true} –>
<ol class=”wp-block-list”><!– wp:list-item –>
<li>Run <code>/model</code> in a session to confirm which model and effort level you’re actually on.</li>
<!– /wp:list-item –><!– wp:list-item –>
<li>Some early reports suggest Opus 5.5 can default to a lower effort level than your <code>settings.json</code> requests, until you add an explicit entry for the model under <code>modelSettings</code>. That claim comes from secondary coverage rather than the official changelog, so check your own effort setting rather than assume it.</li>
<!– /wp:list-item –><!– wp:list-item –>
<li>If you rely on MCP servers with long tool descriptions, 2.1.280 added an environment variable to raise the previous 2,048-character cap.</li>
<!– /wp:list-item –><!– wp:list-item –>
<li>Watch your token usage for a day or two. A bigger window makes it easy to keep far more loaded in a session, and that has a cost even at the lower per-token price.</li>
<!– /wp:list-item –></ol>
<!– /wp:list –>

<!– wp:heading –>
<h2 class=”wp-block-heading”>Where the bigger window actually helps</h2>
<!– /wp:heading –>

<!– wp:paragraph –>
<p>With a 1M-token window you can load a whole module plus its tests up front and ask for a change that touches all of it, rather than feeding files in piecemeal. That’s a real advantage for migrations, API renames, and test-suite clean-ups.</p>
<!– /wp:paragraph –>

<!– wp:paragraph –>
<p>It doesn’t remove the need for clear direction, though. A concise project instructions file and a well-scoped task description still beat dumping an entire repository in and hoping for the best – extra context is capacity, not a substitute for telling the model what you actually want.</p>
<!– /wp:paragraph –>

<!– wp:heading –>
<h2 class=”wp-block-heading”>Should you switch?</h2>
<!– /wp:heading –>

<!– wp:paragraph –>
<p>If you were already defaulting to Opus, you’ve switched automatically and your cost per task should drop. If you’d moved to Sonnet purely for cost reasons, it’s worth running one real task on Opus 5.5 and comparing the bill and the result before deciding anything permanent – Sonnet 5.5 is due soon and may shift that calculation again. Either way, check Anthropic’s own release notes for the authoritative details, since several specifics here come from coverage published within days of launch.</p>
<!– /wp:paragraph –>