Sonnet 5 narrows the gap between a daily assistant and an agent

Anthropic introduced Claude Sonnet 5, emphasizing planning, browser and terminal use, and tasks requiring less manual supervision. The company compares its agentic performance with larger Claude models and positions it for wider adoption.

Context

The release invites users to reassess which operations need their most expensive model. Autonomy is still easiest to evaluate on bounded tasks with inspectable results and reversible changes. A successful short example does not show how reliably an agent handles a long process with shifting context and imperfect tool responses.