Claude Opus 5.5 Explained: Which Plan Gets It and What Changes

Anthropic has shipped a new top-tier model that costs paying subscribers nothing extra, and for once the interesting part is not the benchmark chart. On September 22, 2026 Anthropic released Claude Opus 5.5, "the first model in our new Claude 5.5 family," and paired it with something people on a $20 plan can actually feel: bigger usage limits and a reset you can save for a bad day. The consumer questions are the usual three. Do you get it, does it cost more, and does anything about your Claude change.

Key Takeaways

• Claude Opus 5.5 is available today, September 22, 2026, and Anthropic's help centre release notes carry the launch entry under that date.
• The free plan does not get it. Opus has never been a free-plan model. Free Claude stays on Sonnet 5 and Haiku.
• On Pro ($20/month) it is included at no extra cost, and on both Max tiers too. Claude's pricing page marks the Opus row Yes for Pro, Max 5x and Max 20x, and No for Free.
• Your usage limits go up. Anthropic says it is increasing five-hour limits on Pro, Max, Team and seat-based Enterprise, and giving subscribers a limit reset you can save and use whenever you choose.
• It matches Anthropic's most expensive model on most work. Anthropic says Opus 5.5 "performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5."
• It knows more than Opus 5 did. Training data runs to June 2026, a month past Opus 5's May 2026 and level with Fable 5.1.
• You cannot turn its thinking off. On Opus 5.5 thinking is always on at every effort level, including on the API.
• It watermarks its text. Anthropic's marking table added Opus 5.5 on September 22, 2026, with text watermarks on Anthropic's own surfaces and on AWS, Google Cloud and Microsoft Foundry.
• Developers pay less: $4 per million input tokens and $20 per million output, 20% below Opus 5, with cache reads down 60% to $0.20.

Do You Get It? Availability by Plan

Here is the picture as of September 22, 2026, read off Claude's pricing page plan matrix and Anthropic's announcement.

The row that trips people up is Fable, not Opus. On Pro, Anthropic's top Fable model still bills against pay-as-you-go usage credits rather than your subscription, and Max 5x and Max 20x get Fable up to 50% of weekly limits. Opus is the tier that is simply included. That is why Opus 5.5 matters more to a Pro subscriber than a Fable release does: it is the ceiling of what your $20 buys, and that ceiling just moved.

What Actually Changes If You Pay for Claude

Three things, and only one of them is the model.

Your limits get bigger. Anthropic says it is "increasing five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans." This is the change most likely to show up in your week, because hitting the five-hour wall mid-task is the complaint that actually drives people to cancel.

You get a usage reset you can bank. Anthropic is "providing subscription users a rate limit reset, which you can now save and use whenever you choose." Its help centre article explains the mechanics: a reset sets either your five-hour session limit or your weekly limit back to full the moment you spend it, you do not have to wait until you are blocked, and it is one-way once used. Find it at Settings > Usage, in the Resets section, under a "Reset for free" button. Two caveats worth knowing: an unused reset can expire, with the date shown in Settings > Usage, and the button is not on Claude Mobile or in Claude Code in your terminal, so open Claude in a browser or the desktop app to spend it.

The model is better and quicker. Anthropic says Opus 5.5 generates output "more than 30% faster than Opus 5" and, on its own benchmark table, beats both Opus 5 and Fable 5.1 on every row it publishes. It also flags its own caveat honestly: "at these levels of capability we've found that benchmark margins have become a less reliable guide to real-world differences," and "the gap between Opus 5.5 and Claude Fable 5.1 is narrower than these scores suggest."

What It Knows, and How Much It Can Hold

Two numbers a lot of people want and few articles print. Both come from Anthropic's help centre, read September 22, 2026.

A June 2026 cutoff means Opus 5.5 has no built-in knowledge of anything from the last three months, this article included. For recent events it still has to search the web, exactly like every other model. Treat the cutoff as the date its memory stops, not the date its usefulness stops.

Claude May Still Swap Models on You Mid-Chat

It does, and Opus 5.5 adds a new reason. Anthropic's help article on the behaviour was retitled on launch day to cover both Opus 5 and Opus 5.5, and the rules differ between them.

The detail that catches people out is unchanged from Fable: the checks "review everything the model reads, not just your latest message. This includes memory, content from connectors, web search results, and files, so a fallback can be triggered by content you didn't type."

For ordinary use none of this fires. Anthropic is explicit that you can still use Opus 5.5 for "everyday health and educational questions, including interpreting lab results, understanding symptoms, and learning about biology," and for secure coding including scanning your own source for vulnerabilities.

Thinking Is Always On, but Effort Is Yours

You cannot switch thinking off, and that is deliberate. Anthropic's help article on model settings says thinking "cannot be turned off in Claude when using Claude Opus 5.5, Claude Fable 5.1, or Claude Opus 5," and that on Opus 5.5 "thinking is always on at every effort level, including on the Claude API."

What you do control is effort, from the same menu next to the send button. Low and Medium "work well for routine tasks and stretch your usage further," High is the recommended balance, and Max is for the deepest analysis. If your usage limit is the thing you care about, effort is the dial that moves it, not the thinking toggle. Higher effort uses more tokens and gets you to the wall faster.

Yes, It Watermarks What It Writes

Text written by Opus 5.5 carries an invisible watermark, from day one. Anthropic revised its help centre article on marking AI-generated content on September 22, 2026 to add an Opus 5.5 row, ticked for text watermarks on Anthropic's own surfaces, for text watermarks through AWS, Google Cloud and Microsoft Foundry, and for Content Credentials on generated files.

This follows Anthropic's own rule that "Claude models launched in the EU on or after August 2, 2026 will support machine-readable marking at launch." The mark carries no information about you, costs nothing, and only Anthropic's private-preview detection API can read it. Our full explainer on Claude's text watermark covers what it can and cannot prove.

Should You Change Anything?

No, if you are on free Claude. Opus has never been on the free plan and 5.5 does not change that. Nothing about your Claude moved today. If you are weighing whether to start paying for something, which AI chatbot is worth paying for walks through the comparison.

Nothing to do, if you are on Pro, but check your model picker. Opus 5.5 is included at your existing price and it is now the best model your $20 covers. Click the model name next to the send button and select it. Your five-hour limit should also be larger than it was yesterday.

Go and look at Settings > Usage, on any paid plan. The saveable limit reset is the genuinely new mechanic here, and it is the kind of thing people discover three weeks late, after the week they needed it. Check whether yours has an expiry date attached.

Do not upgrade from Pro to Max because of this release. Opus 5.5 is included on both. Max buys you 5x or 20x the usage, priority access at busy times, and Fable up to 50% of weekly limits. Those reasons are exactly what they were last week.

If you pay per token, re-run your numbers. Input and output are 20% cheaper than Opus 5, cache reads are 60% cheaper at $0.20 per million, and Anthropic estimates 40% off typical workloads once you account for it using fewer tokens per task.

The bottom line: this is the rare model release where the plan matrix stays still and the value on the paid tiers goes up anyway. Free Claude is untouched. Pro gets a better included ceiling, a bigger five-hour limit and a reset it can bank, for the same $20. The two things worth carrying away are that thinking can no longer be switched off, and that Claude will quietly move you to a different model mid-chat on a narrow set of topics unless you tell it not to. Sonnet 5.5 and Haiku 5.5 are next, "in the coming weeks."

Sources

• Anthropic, Introducing Claude Opus 5.5, page dated September 22, 2026
• Anthropic, Release notes, help centre, entry dated September 22, 2026: "Claude Opus 5.5 launch"
• Anthropic, claude.com/pricing plan comparison matrix, decoded September 22, 2026
• Anthropic, How up-to-date is Claude's training data?, read September 22, 2026
• Anthropic, How large is the context window on paid Claude plans?, read September 22, 2026
• Anthropic, Change the model, effort, and thinking settings, read September 22, 2026
• Anthropic, Why Claude switched models in your conversation with Opus 5 or Opus 5.5, read September 22, 2026
• Anthropic, What is a limit reset?, read September 22, 2026
• Anthropic, How Claude marks AI-generated content, lastUpdatedDate 2026-09-22T16:28:40Z