Claude Sonnet 5.5 Explained: Free Claude Gets the New Model Too

Anthropic has released a new Claude model, and for once the free plan is not the afterthought. On September 28, 2026 it launched Claude Sonnet 5.5, the second model in the Claude 5.5 family, six days after Opus 5.5. The difference that matters to most people is one sentence on Anthropic's own model page: anyone can chat with Claude using Sonnet 5.5 on Claude.ai. Opus 5.5 and Fable 5.1 do not get that sentence. This one does.

Key Takeaways

• It is out today, September 28, 2026, and Anthropic's help centre release notes carry the launch entry under that date.
• Free Claude gets it. Anthropic's Sonnet page says "anyone can chat with Claude using Sonnet 5.5 on Claude.ai, available on web, iOS, and Android." Its Opus and Fable pages instead list Pro, Max, Team and Enterprise, so the wording is a real distinction, not filler.
• Anthropic has not called it the default. It said exactly that about Sonnet 5 in June. This time it did not, so assume you have to pick it from the model menu.
• It knows five months more than Sonnet 5. Training data runs to June 2026 against Sonnet 5's January 2026, level with Opus 5.5 and Fable 5.1.
• It is faster. Anthropic says Sonnet 5.5 generates output "30%+ faster than Sonnet 5," making it the fastest Sonnet so far.
• It is the first Sonnet that watermarks its text. Anthropic's marking table added a Sonnet 5.5 row today, ticked in all three columns. Sonnet 5 still marks files only.
• It is also the first Sonnet that can swap itself out mid-chat, falling back to Sonnet 5 on a narrow set of security topics.
• Nothing about your bill changes. Sonnet is included on every plan and there is no announced limit change, unlike the Opus 5.5 release.

Do You Get It? Availability by Plan

Here is the picture as of September 28, 2026, read off Anthropic's Sonnet model page and the claude.com/pricing plan matrix.

Why we can say Free gets it, when Anthropic never writes the word "free". Anthropic's four model pages use the same "Availability and pricing" heading and then split into two wordings. Sonnet 5.5 and Haiku 4.5 get "anyone can chat with Claude using [model] on Claude.ai." Opus 5.5 and Fable 5.1 instead get "available to Pro, Max, Team, and Enterprise users." The company is drawing the line itself, in the same sentence slot, on the same template. The pricing page agrees: the Sonnet row is ticked for Free, Pro, Max 5x and Max 20x, and the Opus row reads No for Free.

The One Thing Anthropic Did Not Say

It did not say Sonnet 5.5 is your new default, and that is a change in behaviour worth noticing. When Sonnet 5 shipped on June 30, 2026, the announcement said plainly that it was "the default model for its Free and Pro plans", and free users woke up on the new model without touching anything. Today's announcement, the model page and the help centre release note all say nothing about defaults.

There is a hint in the other direction, too. Anthropic's new help article on model switching says that after an automatic fallback "the model picker stays on Sonnet 5 for the rest of the chat," which only makes sense if Sonnet 5 is still a model you can sit on.

So do this: open Claude, click the model name next to the send button, and see whether Sonnet 5.5 is listed and whether it is already selected. If it is not there yet, that is normal for launch day and it is not something you can hurry. We will update this page when Anthropic says anything about defaults.

What It Knows, and How Much It Can Hold

Two numbers people want and few launch articles print. Both come from Anthropic's help centre, read September 28, 2026, hours after it was revised for this launch.

The jump from January to June 2026 is the largest single upgrade here for a free-plan user, because it is the one you notice without benchmarks. Sonnet 5 stops knowing things in January; Sonnet 5.5 carries five more months before it has to go and search.

One honest gap: Anthropic's context-window article is titled and scoped "on paid Claude plans." It does not state a number for the free plan, and no other Anthropic page does either, so we are not printing one.

Claude Can Now Swap Models on You Mid-Chat

This is new for the Sonnet line, and free-plan users have not met it before. Anthropic's help article on the behaviour went live today and explains why: "Because its cybersecurity capabilities are a substantial step up from Claude Sonnet 5's, Claude Sonnet 5.5 is the first Sonnet model with cyber fallbacks."

For ordinary use none of this fires. Anthropic says "routine software development and most life sciences work are unaffected," you can still scan your own source code for vulnerabilities, and conversational questions like "why did you do that?" are explicitly not affected.

The detail that catches people out: 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 or block can be triggered by content you didn't type."

You can turn the switching off. Go to Settings > Capabilities and toggle Switch models when a message is flagged off. A flagged request then pauses the chat instead of quietly changing model, which is the better setting if you would rather know.

Yes, It Watermarks What It Writes

Text written by Sonnet 5.5 carries an invisible watermark from day one, and this is the first time that has been true of a Sonnet model. Anthropic revised its help centre article on marking AI-generated content today to add a Sonnet 5.5 row ticked in all three columns: text watermarks on Anthropic's own surfaces, text watermarks through AWS, Google Cloud and Microsoft Foundry, and Content Credentials on generated files.

Sonnet 5 still does not watermark text. In the same table it is ticked for file Content Credentials only, as are Haiku 4.5 and every model released before August 2, 2026. Anthropic's stated deadline for those has not moved: it is "adding watermarks to outputs from models released before August 2, 2026, with all covered by December 2, 2026." Our explainer on Claude's text watermark covers what a mark can and cannot prove.

Thinking Is Always On, Effort Is Yours

You cannot switch thinking off on Sonnet 5.5 in the Claude apps. Anthropic's help article on model settings, also revised today, now lists Sonnet 5.5 alongside Opus 5.5, Fable 5.1 and Opus 5 as models where "thinking cannot be turned off in Claude."

What you control is effort, from the same menu next to the send button. The default in the Claude apps is Medium. Low and Medium "work well for routine tasks and stretch your usage further," High is the recommended balance, and Max is the most thorough and the slowest. If usage limits are what you keep running into, effort is the dial that moves them.

Should You Change Anything?

Yes, if you are on free Claude: go and look at your model menu. This is the rare release where the new model reaches the free tier on day one, and it is a real upgrade rather than a renaming. Five more months of knowledge and noticeably faster answers, for nothing.

Nothing to do on Pro or Max. Sonnet has always been included. Your limits were raised six days ago with Opus 5.5; nothing in today's release changes them again.

Do not start paying because of this. Everything new today is on the plan you already have. If you are weighing a subscription for other reasons, which AI chatbot is worth paying for works through it.

If you write about security or biology, know about the fallback before it surprises you, and decide whether you want the switching toggle on or off.

If you pay per token, the list price did not move: $2 per million input and $10 per million output, the same as Sonnet 5. Anthropic's claim is that it needs fewer tokens for the same work, "up to 30% less per task."

The bottom line: free Claude just got the newest model in the family on launch day, which is not how Opus or Fable releases work and not how most of the industry works either. The two things to carry away are that you may have to select it yourself, and that Claude will now quietly move you to Sonnet 5 on a narrow set of topics unless you tell it not to. Haiku 5.5 is still to come, "in the coming weeks."

Sources

• Anthropic, Introducing Claude Sonnet 5.5, page dated September 28, 2026
• Anthropic, Claude Sonnet model page, "Availability and pricing," read September 28, 2026
• Anthropic, Release notes, help centre, entry dated September 28, 2026: "Claude Sonnet 5.5 launch"
• Anthropic, claude.com/pricing plan comparison matrix, decoded September 28, 2026
• Anthropic, Why Claude switched models in your conversation with Sonnet 5.5, created September 28, 2026
• Anthropic, How up-to-date is Claude's training data?, revised September 28, 2026
• Anthropic, How large is the context window on paid Claude plans?, revised September 28, 2026
• Anthropic, Change the model, effort, and thinking settings, revised September 28, 2026
• Anthropic, How Claude marks AI-generated content, lastUpdatedDate 2026-09-28T18:34:34Z