The Claude Opus 5.5 Switch Guide

Anthropic released Claude Opus 5.5 on September 22. It performs at the level of Fable 5.1 on most work and costs 40% less to run than Opus 5. Here is how to switch your agents over in five steps, which settings matter, and the audit brief I would hand it first.

01Step 01

Switch your default model in two minutes

In a terminal, run claude update (Opus 5.5 needs Claude Code v2.1.280 or later). Then inside Claude Code type /model opus. On the Anthropic API the opus alias now points to Opus 5.5. Building on the API instead? Change the model string to claude-opus-5-5. Anthropic's own docs now say: if you are unsure which model to use, start with Opus 5.5.

02Step 02

Pick the effort level per job, not per account

Opus 5.5 defaults to medium effort, the cheaper setting that suits most work. Run /effort in Claude Code to open the slider, or /effort high for a hard task. On the API, set output_config.effort. My rule: medium for everyday agent runs, high or xhigh for a migration or an audit, max only when a result keeps falling short.

03Step 03

Cut agent costs with caching first

Pricing is $4 input and $20 output per million tokens, 20% below Opus 5. The bigger drop is cache reads: $0.20 per million, 60% less. Anthropic says cache reads make up most of the cost of agentic and coding work. So put everything that never changes at the top of your agent's prompt (role, rules, tool docs, examples) and only the new task at the bottom. That stable top part is what gets cached.

04Step 04

Hand it the long, sprawling job

This is where Opus 5.5 pulls ahead. Anthropic reports an early tester audited and fixed a 200,000-line codebase in under three hours, where Opus 5 took over 20. Paste the audit brief below into Claude Code in your project folder, at high effort. It lists problems before it touches anything, so you approve the fixes. This reel's research, script and edit ran in one Claude Code session on Opus 5.5.

05Step 05

Re-run your client agents and show the difference

If you sell AI systems, this is a free upgrade to hand your clients. Pick one agent you already deliver, run its usual week of tasks on Opus 5.5 at medium effort and note the time and token cost next to last week's. When a client needs speed over cost, fast mode runs up to 2.5x faster in Claude Code and on the Claude Platform. It is billed at a higher rate, so keep it for urgent runs.

Audit brief · paste into Claude Code in your project folder

# AUDIT BRIEF · run at /effort high
Role: you audit this codebase before we change anything.
1. Map the project: entry points, main modules, dependencies.
2. List the 10 highest-risk problems: bugs, security holes,
dead code, slow paths. File and line for each.
3. Rank them by impact vs effort in one table.
4. Propose a fix plan for the top 3. Do NOT edit files yet.
5. Wait for my OK, then fix one item at a time and run
the tests after each fix.
Done when: the top 3 are fixed and every test passes.

Want this running for you?

Build it step by step inside AI Creative OS, with me in the thread.

Get Your First Agent →

Not ready for that?

Drop your email and 3 more guides open in the Library, free.

Unlock all 84 guides

Get Your First Agent →