Anthropic released Claude Sonnet 5 on Tuesday, a midsize model the company says can make plans, use tools like browsers and terminals, and run autonomously at a level that previously required larger, more expensive models. The introductory price is $2 per million input tokens and $10 per million output tokens through August 31, 2026, then rises to $3 and $15.
Agentic abilities are now considered a standard requirement across all pricing levels, according to Anthropic. Sonnet 5 delivers results nearly as good as Opus 4.8 at a fraction of the expense. As of Tuesday, it becomes the standard offering in both free and Pro tiers and can be accessed across all subscription options.
Sonnet 5 is cheaper than Opus 4.8, as well as OpenAI's GPT-5.5 and Google's Gemini 3.1 Pro. (It is still more expensive than Gemini 3.5 Flash.) Zapier senior engineer Daniel Shepard tested Sonnet 5 on a two-part job.
"We handed Claude Sonnet 5 a two-part job - update Salesforce account tiers, send a launch announcement to enterprise contacts - and it finished end to end," Shepard said. "That used to stall halfway. For day-to-day automation, it's a no-brainer."
Get your free investing masterclass bonus when you join Market Briefs, our free daily newsletter
Anthropic reports that Sonnet 5 shows major gains relative to Sonnet 4.6 in areas such as reasoning, tool usage, coding, and knowledge tasks.
On a benchmark for agentic coding, Sonnet 5 posted 63.2%; Opus 4.8 scored 69.2% and Sonnet 4.6 scored 58.1%. For knowledge work tasks, Sonnet 5 edged past Opus 4.8. Anthropic said in a blog post: "Opus 4.8 is still the model of choice for higher accuracy on these tasks, but Sonnet 5 provides developers with lower-priced options that are of much higher quality than what was previously available."
On safety, Sonnet 5 shows fewer "undesirable behaviors" such as collaborating with misuse or being deceptive compared to its earlier version. It exhibits improved ability to decline harmful requests and avoid hijacking via prompt injections. Additionally, it hallucinates and displays sycophantic behavior less frequently than Sonnet 4.6.
Nevertheless, it falls short of Opus 4.8 and Claude Mythos Preview in terms of resisting misaligned behavior. Anthropic also says evaluations show Sonnet 5 has a much lower ability to perform dangerous cybersecurity tasks than current Opus models.
According to Lovable co-founder Fabian Hedin, "A model that knows when to say no is just as important as one that knows how to build."
The price jump after August 31 will push some customers to decide whether the extra accuracy of Opus 4.8 is worth the cost. Meanwhile, OpenAI released GPT-5.6 Sol in preview the week prior, and Google launched Gemini 3.5 Flash in May.
Subscribe to Market Briefs, our free daily newsletter, and claim your bonus investing masterclass
