Anthropic ships Claude Sonnet 5
Sonnet 5 is now the free and Pro default at $2/$10 per million tokens intro pricing, with agentic coding at 63.2%, up from Sonnet 4.6's 58.1%.
Anthropic launched Claude Sonnet 5 on June 30 and made it the default model for free and Pro plans starting July 1. Intro pricing is $2 per million input tokens and $10 per million output tokens through August 31, after which it moves to $3 and $15.
On the agentic coding benchmark, Sonnet 5 scores 63.2%, up from Sonnet 4.6’s 58.1% and closing most of the distance to Opus 4.8’s 69.2%. It is available on Free, Pro, Max, Team, and Enterprise plans and through the API as claude-sonnet-5.
Two changes that affect builders
First, the model checks its own output without being asked. In practice this reduces truncation and mid-task stalls in multi-step pipelines, which has been a friction point with Sonnet 4.6. Second, an updated tokenizer consumes 1.0 to 1.35x more tokens per request than its predecessor. That is the reason for the intro pricing period, and it is also why the cost gap against Opus 4.8 narrows on verbose outputs. Run your actual workload, not just benchmark arithmetic, before committing to a budget.
Hallucination and sycophancy rates are lower. Cybersecurity exploit capabilities are intentionally capped below Sonnet 4.6, with cyber safeguards enabled by default.
What to do now
The clearest action is to rerun your eval suite on Sonnet 5. For many agentic workflows this may remove the need for Opus 4.8 entirely. The rate limit increases Anthropic rolled out earlier this month mean Sonnet 5 is also less likely to hit ceiling during load spikes. If you are deciding where Sonnet 5 sits in your stack, the guide on choosing reasoning depth gives a practical framework, though Sonnet 5 shifts the numbers in that analysis meaningfully.