Claude Sonnet 5 Is Born! Opus 4.8-Level Power at Sonnet's Price!

Hey everyone, it's Shii-chan! I've brought you something exciting today: Anthropic has released a new model, Claude Sonnet 5! And it delivers performance approaching the higher-end Opus 4.8 at Sonnet's friendly price. I can't stop being excited, so let me tell you all about it!

Anthropic News anthropic.com

What was announced?

This news comes from Anthropic's News, their channel for launch and update announcements. The star is a new model called Claude Sonnet 5, which Anthropic built to be the most agentic Sonnet model yet.

It can plan on its own, use tools like browsers and terminals, and run tasks autonomously. Compared to its predecessor, Sonnet 4.6, it's a solid step up on the parts of agentic performance that matter: reasoning, tool use, coding, and knowledge work.

Why it matters

For many developers, the agentic AI era began with Sonnet-class models. Claude Sonnet 3.5, 3.6, and 3.7 were the first to show impressive skills in coding and tool use.

More recently, though, the clearest gains in agentic capability have shown up in the Opus-class models, and Sonnet 4.6 fell well short of Opus 4.8.

Claude Sonnet 5 narrows that gap. Its performance is close to Opus 4.8, but at a lower price — a welcome challenge to the assumption that top capability has to be expensive!

What changes

The best part is that you can use it right away! Claude Sonnet 5 is the default model on the Free and Pro plans, and it's available to Max, Team, and Enterprise users. It runs in Claude Code and on the Claude Platform, and developers can call it via the API as claude-sonnet-5.

You can also adjust the effort level between Sonnet 5 and Opus 4.8, so you choose the right balance of cost and performance yourself. Let it think hard on heavy tasks, and keep things efficient on lighter ones!

Dive Deep

Let's look at the details a bit more closely!

Pricing and tokens

Introductory launch pricing runs through August 31, 2026: $2 per million input tokens and $10 per million output tokens. After that, standard pricing is $3 input and $15 output per million tokens. For reference, Opus 4.8 is $5 input and $25 output, so you can see how affordable this is.

Sonnet 5 uses an updated tokenizer that changes how it processes text, so the same input can map to more tokens — roughly 1.0-1.35x depending on the content type. The introductory pricing is set so the move from Sonnet 4.6 is roughly cost-neutral.

How performance looks

The published charts compare Sonnet 5 on the agentic search evaluation BrowseComp and the computer-use evaluation OSWorld-Verified. Sonnet 5 is a clear improvement over Sonnet 4.6, with much better cost efficiency at medium effort, and its higher-effort runs can match Opus 4.8 on some tasks.

Safety

Pre-deployment safety evaluations found Sonnet 5 to be an overall improvement on Sonnet 4.6. It's better at refusing malicious requests and resisting prompt injection, and it shows lower rates of hallucination and sycophancy. That said, its rate of misaligned behavior was somewhat higher than the more capable Opus 4.8 and Mythos Preview.

On cyber tasks, its ability to develop dangerous exploits is much lower than Opus 4.8 and Mythos 5. In a Firefox browser exploit evaluation built with Mozilla, neither Sonnet model could produce a working exploit (both scored 0.0%). Even so, because it's a bit stronger than before, it ships with cyber safeguards enabled by default.

Where you can use it

Beyond the native Claude Platform, it's available on the Claude Platform on AWS and in Claude in Microsoft Foundry (hosted on Azure and Anthropic), and it's coming soon to Claude in Google Vertex.

Wrap-up

• Claude Sonnet 5 is the most agentic Sonnet yet, stronger at planning, tool use, and autonomous execution
• Performance approaches Opus 4.8 while keeping Sonnet's friendly price, and you can tune the effort level to balance cost and performance
• Introductory pricing through August 31 is $2 input and $10 output per million tokens (then $3 and $15)
• It's the default model on Free and Pro, available via the API as claude-sonnet-5, on AWS and Microsoft Foundry, with Google Vertex coming soon
• Safety improved over Sonnet 4.6, dangerous cyber skills stay low, and safeguards are on by default

If you've been thinking, "I want to run agents, but Opus is a little pricey," this new model is a perfect fit!