Anthropic released Claude Sonnet 5 on Tuesday, making it the default model for Free and Pro plans and pricing it at $2 per million input tokens and $10 per million output tokens through August 31. The model is aimed at everyday agent tasks, where cost and autonomy matter at the same time.
Claude Sonnet 5 pricing
The launch gives developers a cheaper entry point for tool-using work than Opus 4.8, OpenAI, and Google models in the same family of use cases, while still leaving a price gap above Gemini 3.5 Flash. Anthropic said the rate will move to $3 per million input tokens and $10 per million output tokens after August 31, so anyone building on the cheaper window has a short-term cost advantage.
Daniel Shepard at Zapier
Daniel Shepard, a senior engineer at Zapier, said Claude Sonnet 5 handled a two-part job end to end. He said the task involved updating Salesforce account tiers and sending a launch announcement to enterprise contacts.
“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.”
Agentic gains over Sonnet 4.6
Anthropic said Claude Sonnet 5 can make plans, use browsers and terminals, and run autonomously at a level that previously required larger and more expensive models. It also said the model improves on Sonnet 4.6 in reasoning, tool use, software coding, and knowledge work, and testers found that it checks its own output without being asked.
The numbers complicate the pitch. Claude Sonnet 5 scores 63.2% on one agentic coding benchmark, while Opus 4.8 scores 69.2% and Sonnet 4.6 scores 58.1%. Anthropic also said Opus 4.8 remains the better choice for higher accuracy on those tasks, even as Sonnet 5 gives developers a lower-priced option.
After August 31 pricing
For a builder choosing between cost and accuracy, the immediate decision is whether the cheaper rate through August 31 is enough to justify moving agent workflows now. The open question is the benchmark method behind the 63.2% score, which Anthropic did not explain in the material behind this launch.
