Claude Sonnet 5.5

Published frontier peer. Closed API claude-sonnet-5-5. Index Terminal-Bench 2.1, SWE-bench Verified, and AutomationBench stay empty because the Anthropic launch table does not publish those LocalsOnly Index versions (Terminal-Bench on the card is 4.0, not 2.1; there is no AutomationBench line; there is no SWE-bench Verified line). Not a LocalsOnly desk run. Launch-table figures that are not Index columns: Terminal-Bench 4.0 70.6% (Sonnet 5 was 10.3%), FrontierCode 1.1 Main 46.2% Max / 52.1% Xhigh, CursorBench 4.0 55.5%, OSWorld 2.1 80.1% partial (not OSWorld 2.0), GDPval-AA v2.1 1844. Pricing matches Sonnet 5 at $2 / $10 per million input / output tokens; up to about 30% less cost per task and 30%+ faster. Complements Opus 5.5 for well-scoped everyday coding. https://www.anthropic.com/claude-sonnet-5-5