What is Claude Sonnet 5?

Claude Sonnet 5 is Anthropic's new default model for Free and Pro plans, released June 30, 2026. It's built for agentic work — coding, tool use, multi-step automation — and performs close to Opus 4.8 at a much lower price.

How much does Claude Sonnet 5 cost?

Introductory pricing through August 31, 2026 is $2 per million input tokens and $10 per million output tokens. After that it moves to standard pricing of $3 per million input and $15 per million output — the same list price as Sonnet 4.6.

Is Claude Sonnet 5 as good as Opus 4.8?

Sonnet 5 gets close on most benchmarks and even beats Opus 4.8 on Terminal-Bench 2.1 and knowledge work (GDPval-AA v2). Opus 4.8 still leads on the hardest reasoning tasks and remains the recommended model for cybersecurity work requiring reduced guardrails.

Does Claude Sonnet 5 use more tokens than Sonnet 4.6?

Sonnet 5 uses a new tokenizer that can produce roughly 1.0 to 1.35 times more tokens for the same input, depending on content type — code and non-English text see the biggest increase. Budget for this, especially once intro pricing ends September 1.

Claude Sonnet 5 Is Live: Near-Opus Power, Way Less Cost

Anthropic shipped Claude Sonnet 5 on June 30. It's now the default model for every Free and Pro user, and it gets close enough to Opus 4.8 that Opus is starting to feel optional for a lot of workloads.

Here's what actually changed and whether you should flip your API calls over.

The Numbers

Sonnet 5 hits 63.2% on agentic coding (SWE-bench Pro) versus Opus 4.8's 69.2% and Sonnet 4.6's 58.1%. On Terminal-Bench 2.1 it actually beats Opus 4.8 — 80.4% versus 74.6%. On knowledge work (GDPval-AA v2) it edges Opus out too, 1,618 to 1,615.

It's the first Sonnet to beat its own Opus sibling on a major benchmark. That's not marketing spin — it's a real shift in where the mid-tier model sits.

Pricing

Introductory price through August 31: $2 per million input tokens, $10 per million output. After that it jumps to $3/$15 — same sticker as Sonnet 4.6. Opus 4.8 is $5/$25. Fable 5 is $10/$50.

One catch worth knowing before you migrate: Sonnet 5 uses a new tokenizer. Same input, roughly 1.0–1.35x more tokens depending on content — code and non-English text take the biggest hit. Anthropic priced the intro rate low enough to offset that, but the offset expires September 1 along with the intro pricing.

Run your own workload through it before you commit budget past August. "Cost-neutral" has an expiration date here.

What's Actually New

Sonnet 5 is built to finish multi-step agentic tasks instead of stalling halfway — a Zapier engineer described handing it a two-part job (update Salesforce, send a launch email) and having it complete both ends without prompting. That used to require Opus.

It ships with a 1M-token context window at no long-context premium, up to 128K output tokens, adaptive thinking on by default, and effort levels from low to xhigh so you can dial cost against accuracy per request.

It's also safer by Anthropic's own numbers: prompt-injection attack success in browser-use scenarios drops from 31.5% to 0.93% versus Opus 4.8 without safeguards, plus lower hallucination and sycophancy rates than Sonnet 4.6.

One place it deliberately stays weaker: cybersecurity. Sonnet 5 never developed a working exploit in Anthropic's Firefox 147 testing, and Anthropic still recommends Opus 4.8 for reduced-guardrail security work.

Should You Switch?

If you're running Claude Code, tool-chaining pipelines, or anything that previously needed Opus 4.8's extra horsepower just to finish reliably, test Sonnet 5 at the intro price. The cost-performance curve moved enough that it's worth the hour it takes to benchmark on your own tasks.

For pure correctness on the hardest reasoning tasks, Opus 4.8 still wins. For everything else — day-to-day coding, automation, knowledge work — Sonnet 5 is the new default for a reason.

It's available now in Claude Code, on the Claude Platform, and across Chat and Cowork. Model ID is claude-sonnet-5.

Sources: Anthropic, TechCrunch