The numbers that matter are on coding and computer use. Sonnet 5.5 scores 70.6% on Terminal-Bench 4.0 — up from Sonnet 5's 10.3% — 55.5% on CursorBench 4.0, and 46.2% on FrontierCode at max effort. On the GDPval agentic benchmark it lands at 1844 Elo, all but level with Opus 5.5's 1846, while output generation runs over 30% faster. Pricing is unchanged from Sonnet 5 at $2 per million input tokens and $10 per million output ($0.20 for cache reads), which is half what Opus 5.5 costs — so the pitch is close-to-Opus results for well-scoped work without the Opus bill. It's the first Sonnet to ship with the cyber safeguards Anthropic previously reserved for its most capable models: cybersecurity capabilities are now rated comparable to Opus 5, so the same fallbacks and classifiers come along, alongside Sonnet 5-level biology protections and defenses against distillation and cross-account capability extraction. The model is live on the Claude Platform (model id claude-sonnet-5-5) and across AWS, Google Cloud, and Azure. Arriving a week after Opus 5.5, it's Anthropic's second 5.5-family release this month — the quieter, cheaper sibling, aimed at everyday coding and document work rather than the frontier.