Claude Sonnet 5.5

Claude Sonnet 5.5 is a mid-tier model in the Claude 5.5 family positioned as a faster, lower-cost complement to the higher-judgment Opus 5.5. It targets well-scoped everyday tasks: bug fixes, polished documents, slides, spreadsheets, design work and high-throughput coding. Sonnet 5.5 runs 30%+ faster than Sonnet 5, typically needs far fewer tokens - yielding up to ~30% lower cost per task - and keeps the same published token pricing ($2 per million input, $10 per million output, $0.20 per million cache reads). It’s optimized for speed and iteration at low-to-medium effort levels while still scaling to higher-effort settings when needed; a Haiku 5.5 variant for cost-sensitive, high-volume use is coming soon.

Benchmarks and user testing show substantial gains: Terminal-Bench 4.0 70.6% (vs 10.3 for Sonnet 5), CursorBench 55.5% (34.1), and GDPval-AA ~1844 (vs 1449), with strong chart/visual recognition and even a first Sonnet success on Pokémon Red from screenshots. Coding improvements include higher FrontierCode scores and fewer tool calls, producing faster multi-file workflows and fewer stalled builds. Alignment audits equal or improve on Sonnet 5; Sonnet 5.5 ships with cybersecurity safeguards comparable to Opus while retaining prior biology safeguards focused on high-risk requests. Enterprise testers report faster iteration, lower token use, better judgment, and near-sendable outputs for slide and support workflows.