Anthropic releases Claude Sonnet 5 with stronger agentic capabilities and lower pricing

Anthropic released Claude Sonnet 5, positioning it as a more powerful and agentic version of its mid-size model. The company describes the model as capable of making plans, using tools like browsers and terminals, and running autonomously at a level previously requiring larger and more expensive models.

The model is priced at $2 per million input tokens and $10 per million output tokens through August 31, after which the price increases to $3 per million input tokens and $15 per million output tokens. Anthropic frames Sonnet 5 as a cheaper alternative to Opus 4.8, OpenAI’s GPT-5.5, and Google’s Gemini 3.1 Pro, though it remains more expensive than Google’s Gemini 3.5 Flash.

On agentic performance benchmarks, Sonnet 5 scores 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 outperforms Opus 4.8. Anthropic notes that Opus 4.8 still leads on complex tasks requiring subtle judgment calls and deep research.

The company highlights improvements in safety, including a lower rate of undesirable behaviors such as cooperation with misuse and deception. Sonnet 5 also shows better resistance to prompt-injection attacks and hallucinates less frequently than Sonnet 4.6. However, Anthropic states that its safety performance does not yet match Opus 4.8 or Claude Mythos Preview.

Anthropic positions Sonnet 5 as a model that balances cost and performance, enabling users to adjust effort levels based on their needs. The company cites testers who report that Sonnet 5 completes complex tasks end-to-end without stalling, including an example where it updated Salesforce account tiers and sent a launch announcement to enterprise contacts.