Anthropic launches Claude Sonnet 5 as a cheaper agentic model

Anthropic launches Claude Sonnet 5 as a cheaper agentic model

9 reported

Anthropic has released Claude Sonnet 5, a midsize model designed for agentic tasks such as planning, using tools like browsers and terminals, and running autonomously. The company stated in a blog post that the model performs at a level that previously required larger and more expensive models. Starting Tuesday, Sonnet 5 will be the default model for free and Pro plans and is available for every subscription. At launch, it is priced at $2 per million input tokens and $10 per million output tokens through August 31, after which the price will rise to $3 per million input tokens and $10 per million output tokens. Anthropic said Sonnet 5 shows significant improvements over its predecessor Sonnet 4.6 on agentic performance benchmarks, including reasoning, tool use, software coding, and knowledge work. The company also reported that Sonnet 5 has a lower rate of undesirable behaviors such as cooperation with misuse and deception compared to Sonnet 4.6.

What’s reported

Claude Sonnet 5 is a midsize model from Anthropic designed for agentic tasks.
It can make plans, use tools like browsers and terminals, and run autonomously.
Starting Tuesday, Sonnet 5 will be the default model for free and Pro plans.
Launch pricing is $2 per million input tokens and $10 per million output tokens through August 31, then $3 per million input tokens and $10 per million output tokens.
Sonnet 5 is cheaper than Opus 4.8, OpenAI’s GPT-5.5, and Google’s Gemini 3.1 Pro, but more expensive than Gemini 3.5 Flash.
On a benchmark, Sonnet 5 scored 63.2% on agentic coding, compared to Opus 4.8’s 69.2% and Sonnet 4.6’s 58.1%.
On a knowledge work benchmark, Sonnet 5 slightly outperformed Opus 4.8.
Sonnet 5 has a lower rate of undesirable behaviors than Sonnet 4.6, including better refusal of malicious requests and lower hallucination rates.
It is not on the same level as Opus 4.8 and Claude Mythos Preview for misaligned behavior.

Key figures

Daniel Shepard, senior engineer at Zapier
Fabian Hedin, co-founder of Lovable

Sources: TechCrunch

You may also like...

Leave a Reply

Your email address will not be published. Required fields are marked *