Claude Sonnet 5 makes agentic work the default tier

Anthropic released Claude Sonnet 5 on June 30 across Claude plans, Claude Code and its API. The company made the model the default for Free and Pro users, moving its newest agent-oriented capabilities into the tier most people encounter first rather than reserving them for a specialist or premium-only product.

Anthropic positioned Sonnet 5 for agentic coding, browser use, planning, tool use and knowledge work. Those labels describe a model intended to carry tasks through several steps, but they do not remove the need for permissions, supervision or review. The practical question for subscribers is which jobs can be delegated reliably and which still need frequent intervention.

Launch pricing was set at $2 per million input tokens and $10 per million output tokens through Aug. 31. Anthropic said pricing would then move to $3 and $15 respectively. Because that transition was scheduled after this retrospective was prepared, developers should check the current rate before estimating production costs.

Anthropic said its pre-deployment safety evaluations found fewer undesirable behaviors than Sonnet 4.6. That remains a provider assessment. The company also edited an evaluation chart on launch day after identifying a methodology mismatch, a reminder that launch graphics can change as methods are reviewed.

Sonnet 5's wider availability may matter more than any single benchmark. Putting an agentic model in the default slot increases the number of people who can try multi-step work without changing plans or model settings. Evidence from ordinary tasks, including completion quality, tool errors, supervision time and total cost, will be more useful than treating one company-run score as a settled ranking.