In This Article
Key Takeaways
- On June 30, 2026, Anthropic released Claude Sonnet 5 and made it the default model for Free and Pro users, with access for Max, Team, and Enterprise plans and inside Claude Code.
- Introductory pricing is $2 per million input tokens and $10 per million output through August 31, 2026, then $3 and $15 after that.
- Anthropic positions Sonnet 5 as its most agentic Sonnet yet, with performance close to Opus 4.8 on many tasks at a lower price.
- The notable caveat: Anthropic states Sonnet 5 has a much lower ability to perform cybersecurity tasks than its current Opus models.
If your default Claude model changed under you at the end of June and you did not notice, that is by design. On June 30, 2026, Anthropic released Claude Sonnet 5 and quietly made it the default across the product. Most people using Claude on a Free or Pro plan are now talking to Sonnet 5 whether or not they went looking for it. This piece is our analysis of what that swap means in practice — not a launch recap, but a read on where it helps, where it does not, and what a working builder should actually do with the news.
What Anthropic actually shipped
The short version: a cheaper model that behaves more like an expensive one. Anthropic describes Sonnet 5 as its most agentic Sonnet to date — a model built to make plans, use tools like browsers and terminals, and run multi-step work on its own with less babysitting. Per Anthropic's platform documentation, it became the default for Free and Pro plans and is available to Max, Team, and Enterprise users, as well as inside Claude Code and through the API as claude-sonnet-5.
The framing that matters is the comparison to Opus. Anthropic says Sonnet 5's performance is close to that of Opus 4.8, and that on some agentic tasks it matches Opus 4.8's capability levels — work that, a few months ago, needed a larger and pricier model. TechCrunch framed the release the same way: a cheaper way to run agents. That is the whole story in one line. The interesting part of a mid-tier model getting more capable is not the benchmark; it is the cost of doing the same job.
The pricing, read plainly
Here is where a default swap turns into real money. Through August 31, 2026, Sonnet 5 runs at an introductory price of $2 per million input tokens and $10 per million output tokens, per Anthropic's own announcement. After that window it moves to standard pricing of $3 input and $15 output. MacRumors reported the same figures on launch day.
Two things follow from those numbers. First, if you are running an agent or a batch job that burns tokens by the millions, the difference between $2 and $3 on input is not a rounding error — it is a 50% swing you can lock in by doing your heavy runs before September. Second, output tokens are five times the price of input tokens, which is the usual shape for these models and a reminder that verbose responses cost more than verbose prompts. If you want to see what any of this does to a real workload, our LLM price calculator and AI cost calculator let you plug in token volumes and compare models side by side before you commit.
The one trade-off to know
A cheaper default is not a free lunch, and Anthropic is unusually direct about the catch. In the same release, the company states that Sonnet 5 has a much lower ability to perform cybersecurity tasks than its current Opus models. That is not a bug — for most teams it is a feature, since a widely deployed default with dialed-down offensive-security capability is a sensible safety posture. But it is a real capability line.
If your work touches security, read the fine print
Anthropic says Sonnet 5's cybersecurity abilities are markedly below its Opus models. For everyday coding, writing, research, and agent workflows, that will not surface. But if you rely on the model for vulnerability analysis, exploit reasoning, or serious security tooling, the default may quietly underperform what you had before — and the fix is to point that specific work at Opus, not to fight the default.
Who should care, and why
Not everyone needs to do anything. If you use Claude as a chat assistant for drafting and questions, the honest answer is that you will mostly notice faster, steadier tool use and nothing to configure. The people who should care are the ones spending money on tokens or shipping on top of the API.
Builders running agents. This is the group the release is aimed at. If you had workloads parked on Opus purely because Sonnet could not hold a multi-step plan together, it is worth re-testing them on Sonnet 5. The pitch is Opus-adjacent behavior at Sonnet prices, and the only way to know if it holds for your task is to run your own evals, not Anthropic's. If you work in Claude Code, the model is already available there — our guide to Claude Code walks through choosing a model per task.
Teams watching the bill. A default that is cheaper and more capable is the rare change that helps cost and quality at once. The move is to route the bulk of your traffic to Sonnet 5 and reserve Opus for the narrow slice of work that genuinely needs it — which, after this release, is a smaller slice than it was. We wrote about that dynamic when Opus 4.8 shipped, in what Opus 4.8 actually means.
Learners choosing what to practice on. If you are getting hands-on with agents and tool use, a stronger, cheaper default lowers the cost of experimenting. That matters more than it sounds — the fastest way to learn this material is to run a lot of small jobs, and cheaper tokens mean more reps for the same budget.
What to do this month
Our practical read comes down to three moves. One: if you run token-heavy jobs, schedule the big ones before August 31 to catch the introductory rate, and model the September price change now rather than being surprised by it. Two: re-benchmark any Opus workloads that exist only because older Sonnet was not agentic enough — some of them will move down a tier cleanly, and some will not, and you want to know which is which on your own data. Three: keep security-sensitive work on Opus, given Anthropic's stated cybersecurity gap.
The larger pattern is the one worth internalizing. The frontier model gets the headlines, but the default model is what actually runs the world's traffic. When the default gets cheaper and more capable in the same release, the quiet economics of everything built on top shift — and that is usually a bigger deal than whatever tops the leaderboard this week.
See what the new pricing does to your workload
Plug your token volumes into our free calculators and compare Sonnet 5, Opus, and rival models side by side — no sign-up, no sales pitch, just the math.
Open the price calculatorSources: Anthropic — Introducing Claude Sonnet 5; Claude Platform Docs — What's new in Sonnet 5; TechCrunch; MacRumors. Analysis and framing by Precision AI Academy.
Common questions
Do I have to switch to Sonnet 5? No — if you are on a Free or Pro plan you are already on it, because Anthropic made it the default. If you need Opus for a specific job, you can still select it on the plans that offer it.
How long does the intro price last? The $2 input / $10 output rate runs through August 31, 2026, after which it becomes $3 input / $15 output, per Anthropic's announcement.
Is Sonnet 5 really as good as Opus? Anthropic says it is close on many tasks and matches Opus 4.8 on some agentic ones, but it explicitly notes lower cybersecurity ability. Treat "close" as a reason to test, not a reason to assume.
Where can I use it? Free and Pro (as default), Max, Team, and Enterprise plans, Claude Code, and the API as claude-sonnet-5.