Claude Sonnet 5 Explained: What's New and What It Means for You

What changed (August 14, 2026): two things. Anthropic published how Claude's text watermark works, confirming that future Claude models will hide an invisible watermark in the text they write, at no extra cost and with no identifying information attached -- it changes nothing about which plan gets Sonnet 5, but it is the first thing about Claude's output itself to change in a while. Separately, the availability table below said its facts were "re-checked ... on July 26, 2026" even though its API row was rewritten on August 12, so the date understated what had been verified and when. Every row was re-read off Claude's pricing page today -- Pro is $20 monthly or $17 on annual billing ($200 up front), Max starts at $100, and Sonnet 5 is still $2 / $10 per million tokens with no expiry -- and the line now carries today's date.

What changed (August 12, 2026): the price rise this page warned about is not happening. The availability table said Sonnet 5's API rate was "intro pricing ... through Aug 31, 2026, rising to $3 / $15 after that." Anthropic has since edited its own Sonnet 5 announcement to say the opposite: "Sonnet 5's introductory pricing of $2/MTok input and $10/MTok output has since been made permanent." Claude's pricing page, re-read the same day, lists Sonnet 5 at $2 / $10 with no expiry date attached. The table row now says that. Every other fact on this page was re-verified against the pricing page today and holds: Sonnet on every plan including Free, Opus from Pro up, Fable billing against usage credits on Pro.

What changed (August 5, 2026): this page said free Claude limits you with "a shorter 5-hour cap and a weekly cap," citing Anthropic's usage-and-length-limits article. That article says neither -- it never mentions hours or weeks, and it is not about the free plan. The weekly cap was not a published Anthropic fact and has been removed. The five-hour half is real, and Anthropic publishes it in a different place: its getting-started article says the free plan has "a session-based usage limit that will reset every five hours" and that message counts "vary based on demand." The FAQ answer and the free-tier section below now say that instead.

What changed (July 24, 2026): Anthropic released Claude Opus 5, which replaces Opus 4.8 as the Opus-tier model and is now the default on Claude Max and the strongest model included with Claude Pro. Nothing changes on the free plan -- Sonnet 5 is still the free default, and everything below still applies. The comparisons to Opus 4.8 on this page are the ones Anthropic made in June 2026, when Opus 4.8 was the flagship.

Here is the short version: Claude just got smarter and you probably did not notice. On June 30, 2026, Anthropic made Claude Sonnet 5 the default model for its Free and Pro plans, replacing the older Sonnet 4.6. So if you open the Claude app or claude.ai today, you are already using it -- no download, no setting to flip, no extra cost. Anthropic said Sonnet 5 performs close to Opus 4.8 -- its most capable model at the time -- and is markedly better at planning and using tools like browsers.

Key Takeaways

• Sonnet 5 is the new default for free and paid Claude users as of June 30, 2026. It replaced Sonnet 4.6.
• Free users get it too. This is the headline for most people: a brand-new, near-top-tier model is now the free default, at no cost.
• You do not need to do anything. The switch was automatic. Open Claude and you are already on Sonnet 5.
• It landed close to Anthropic's best. Anthropic says Sonnet 5 lands near Opus 4.8 in quality while being a "substantial improvement" over Sonnet 4.6. Opus 4.8 has since been superseded by Opus 5.
• It is better at doing things, not just answering. It can make plans and use tools like browsers and terminals more reliably, which matters as Claude takes on multi-step tasks.
• Paid plans get more room and more models, not a different Sonnet -- the model is the same; what you pay for is higher usage limits and access to bigger models, which since July 24, 2026 means Opus 5 on Pro and above.

What Is Claude Sonnet 5?

Claude Sonnet 5 is Anthropic's newest mid-tier model and, as of June 30, 2026, the default you talk to when you use Claude for free or on Pro. "Sonnet" is Anthropic's balanced, everyday line -- fast and cheap enough to run for everyone, but close in quality to the flagship of its day. In Anthropic's words, Sonnet 5 delivers performance "close to that of Opus 4.8, but at lower prices," which is the whole point: the quality that used to require the expensive top model now runs as the standard one. The Opus tier has since moved on to Opus 5, so that gap is wider today than it was in June.

If you have used Claude before and it felt capable, Sonnet 5 is the same experience, quietly upgraded. You did not switch plans or press a button. The model behind the chat box changed.

What's New Compared With Sonnet 4.6?

Two things matter for everyday users: better thinking on multi-step problems, and more reliable use of tools. Anthropic describes Sonnet 5 as more "agentic" -- meaning it can make a plan and then use tools like browsers and terminals to carry it out, running longer stretches of work on its own before checking back with you. If the word "agent" is fuzzy, our plain-terms explainer on what an AI agent actually is covers it in two minutes.

In practice, that shows up as steadier answers on questions that take several steps -- planning a trip, working through a document, pulling live information off the web -- and fewer moments where Claude loses the thread halfway through. Anthropic also raised usage limits across Claude's chat, Cowork, and Claude Code surfaces alongside the launch, so heavier sessions hit the wall a little later than before.

Do You Get It, and on Which Plan?

Yes -- on every plan, free included. Unlike a limited preview, Sonnet 5 shipped to everyone at once as the default. Here is where it stands, re-checked against Claude's pricing page on August 14, 2026:

The one nuance worth understanding: on the free plan, the model is the same as on Pro. What a paid plan buys you is not a better Sonnet -- it is more of it (higher usage limits) plus access to bigger models like Opus 5 and Fable 5, and features like Research and Claude Code. If you are weighing whether that is worth $20, our guide to which AI chatbot is worth paying for breaks it down.

Is It Free, and Where Are the Catches?

It is genuinely free, and the catch is usage, not quality. Free Claude runs Sonnet 5 with no cost, but it caps how much you can use: per Anthropic's help center, the free plan has "a session-based usage limit that will reset every five hours," and "the number of messages you can send will vary based on demand." Anthropic publishes no exact message count, and it reserves the right to "impose other types of usage limits to ensure fair access." What burns through the allowance fastest is length and attachments -- long messages, large pasted documents and files cost more than short chats, per Anthropic's guidance on usage limits. Claude tells you when you have hit the wall.

That makes free Sonnet 5 excellent as a writing and thinking tool you reach for several times a day, and frustrating as an all-day workhorse. If you keep hitting the wall in the same week, that is the real signal it might be time to pay -- see our honest rundown of the best free AI chatbot in 2026 for how Claude's free tier stacks up against ChatGPT and Gemini.

Should You Care?

If you use Claude at all, you already benefited -- so the answer is a low-effort yes. There is nothing to buy and nothing to install; the upgrade already happened. The useful thing to know is simply that free Claude is now meaningfully better than it was in June, which changes the math on whether you need a paid AI subscription at all.

Sonnet 5 also lands in a crowded few weeks. OpenAI's GPT-5.6 Sol family reached general availability on July 9, 2026 -- paid ChatGPT plans get the flagship Sol, and Free and Go are excluded from it -- and Google shipped Gemini 3.1, then Gemini 3.6 Flash on its free tier. If you mostly want a capable everyday assistant, the practical gaps between these are smaller than the headlines suggest -- our ChatGPT vs. Gemini everyday-use comparison walks through what actually matters day to day.

What changed (July 27, 2026): this page still described OpenAI's GPT-5.6 Sol family as "not yet available to regular users," which stopped being true when GPT-5.6 went generally available on July 9, 2026. Corrected. Every Anthropic fact was re-verified the same day against Claude's pricing page and all still hold: Sonnet is ticked on every plan including Free, Opus is included from Pro up, and Fable still bills against usage credits on Pro.

What changed (July 26, 2026): every load-bearing fact on this page was re-verified against Claude's pricing page and Anthropic's Sonnet 5 announcement. All still hold: Sonnet is ticked on every plan including Free, Opus is included from Pro up, and Fable still bills against usage credits on Pro rather than being inside the plan. The availability table's "as of July 4" line was stale after several later edits, so it now carries the re-check date, and the API row now shows the $3 / $15 price the introductory rate rises to after August 31, 2026.

The bottom line: Claude Sonnet 5 is a real, quiet upgrade that reached everyone at once. Free users now run a model that is close to Anthropic's best, at no cost, with no action required. The only reason to pay is if you keep running out of usage or need the bigger models -- not because the free Claude got worse. It got better.