Claude Sonnet 5 brings stronger tool use to business agents

Anthropic released Claude Sonnet 5 on 30 June 2026 and described it as its most agentic Sonnet model so far. The company says the model improves planning, reasoning, coding and the use of tools such as browsers and terminals compared with Sonnet 4.6.

The release is available in Claude, Claude Code and the Claude API. Anthropic also provides selectable effort levels, allowing developers to balance the amount of reasoning, cost and performance for a particular task.

More capable tool use can make a model suitable for longer workflows that combine reading information, making a plan and completing approved actions. A customer operations agent, for example, may need to interpret an enquiry, check a knowledge source, prepare a response and create a structured handover for an employee.

A stronger model does not remove the need for workflow controls. Access permissions, source quality, cost limits, evaluations and human approval still determine whether an agent is dependable in production.

Test the new model against the same real examples and failure cases used for the current system. Compare completed outcomes, tool-call accuracy, latency and total cost instead of relying only on headline benchmark scores.

Anthropic notes that Sonnet 5 uses an updated tokenizer and that identical inputs may use a different number of tokens. Teams should therefore measure usage with their own documents and workflows before estimating production costs.