Introducing Claude Sonnet 5

Details

Claude Sonnet 5 launched June 30 as Anthropic's most agentic Sonnet to date. It closes much of the gap with Opus 4.8 on key benchmarks — including agentic search (BrowseComp) and computer use (OSWorld-Verified) — while being substantially more affordable. The model runs as the default for Claude Free and Pro plans and is available across Max, Team, Enterprise, Claude Code, and the API. On reasoning, tool use, coding, and knowledge work, Sonnet 5 significantly outperforms its predecessor Sonnet 4.6. Anthropic describes Sonnet 5 as covering a wider range of cost-performance options than either Sonnet 4.6 or Opus 4.8: at medium effort it is substantially more cost-efficient than Opus, and at high effort it can match Opus on some tasks. Detailed performance curves are published in the Claude Sonnet 5 System Card. Pricing at launch is $2 per million input tokens and $10 per million output tokens as an introductory offer through August 31, 2026, after which it rises to $3/$15. For reference, Sonnet 4.6 is priced at $3/$15 — meaning Sonnet 5 is cheaper than the outgoing model for the launch period. Safety evaluations found Sonnet 5 shows lower rates of undesirable behaviors than 4.6 and has substantially reduced cybersecurity task capability compared to current Opus models — an intentional design choice given the Fable 5 export control episode. Developers can access the model via the Claude API using the model ID claude-sonnet-5. The introductory pricing window runs through August 31, after which standard rates apply.