Guide · checked against Anthropic's docs on 23 Sep 2026

Switching to Claude Opus 5.5: what changes for you

Start with where you use Claude. It decides more of what changes for you than the model you used before.

New to this? Start here

Opus 5.5 is the newest Claude . If you use it in chat, or , Anthropic has done the technical work and only a few settings change. If your own product calls it through the , a few settings your code sends may now fail or behave differently, and the checklists cover each one.

Using Claude inside another app, like Cursor or GitHub Copilot?

That app decides which models you get and which settings you see. GitHub Copilot is rolling Opus 5.5 out on its paid plans, and on Business and Enterprise your admin controls which models you see. Cursor lists a standard, a thinking and a fast version. Whether a flagged request is handed to another model is up to each app, and neither documents it. The API checklists below are for the company that built the app.

By Anish Gandhi, who builds and rescues production AI apps on the Claude API. Published 23 Sep 2026, the day after Opus 5.5 launched.

Questions people ask

Is Claude Opus 5.5 in the Claude app?

Yes, on Pro, Max, Team and Enterprise plans, where it is first in the model menu. Free plans do not include Opus models. It is also the default model in Claude Code from version 2.1.280.

Can I stop Claude switching to another model?

Yes. In the Claude app and Cowork, open Settings > Capabilities. In Claude Code, run /config. Turn off Switch models when a message is flagged, and a flagged message pauses so you can reword it instead.

Why does Claude Code ignore my effort setting on Opus 5.5?

Claude Code now saves effort per model. A level you saved earlier stays with the model you saved it on, and Opus 5.5 starts at medium. Run /effort while on Opus 5.5 to set its level.

Can I turn thinking off on Opus 5.5?

No. The Claude app and Claude Code have no switch for it, and on the API thinking: {"type": "disabled"} returns a 400 error at every effort level. Lower the effort setting instead.

What is the default effort on Opus 5.5?

Medium on the Claude API and in Claude Code. Opus 5, Opus 4.8 and Fable 5.1 default to high, so a request that never sets effort runs one level lower on Opus 5.5. In the Claude app, the recommended level is marked Default in the effort menu.

What are the context window and knowledge cutoff of Opus 5.5?

A 1M-token context window, up to 128K output tokens per request, and a reliable knowledge cutoff of June 2026.

Sources

Every fact on this page comes from Anthropic's own documentation, checked on 23 Sep 2026, unless it is labelled as a tester's report. Numbers from testers come from one small team with early access. Treat them as early impressions.

Anthropic's documentation

What's new in Claude Opus 5.5 The four breaking changes and the behaviour list
Models overview Latency, default effort, cutoffs, retirement dates
Why Claude switched models Which refusal goes to which model, and the setting that turns switching off
Choosing a model, effort and thinking Effort menu and thinking in the Claude app
Claude Code: model configuration Default model, per-model effort, thinking and flagged requests in Claude Code
Introducing Claude Opus 5.5 Launch post: speed and cost claims