Claude Sonnet 5.5

Claude Sonnet 5.5 is the second model in Anthropic's Claude 5.5 family (after Opus 5.5). It is positioned as a faster, lower-cost complement to Opus 5.5: same per-token price as Sonnet 5 ($2/$10), but Anthropic reports it typically needs far fewer tokens per task, making it up to ~30% cheaper per task in its own testing. It is the fastest Sonnet to date (30%+ faster output generation than Sonnet 5) and adds adaptive thinking with an adjustable effort parameter. Notable improvements are in agentic coding (Terminal-Bench 4.0: 70.6% vs Sonnet 5's 10.3%), knowledge work, and image understanding. Because its cyber capabilities are comparable to Opus 5's, it launched with the same cyber safeguards as Opus 5.5 — a first for a Sonnet-class model. Early enterprise testers (Epic Games, Slack) reported production wins in their own evals. Available via the Claude Platform (claude-sonnet-5-5), AWS Bedrock, Google Cloud, and Microsoft Azure.

Provider-reported (Anthropic, methods in the Sonnet 5.5 system card): Terminal-Bench 4.0 70.6% (Sonnet 5: 10.3%; Opus 5.5: 66.4%); FrontierCode 1.1 Main 46.2% (Max effort); CursorBench 4.0 55.5%; GDPval-AA v2.1 1844 (Sonnet 5: 1449; Opus 5.5: 1846); Humanity's Last Exam 64.5% with tools; OSWorld 2.1 80.1% (partial); Chartography 61.6%.

Independent: enterprise early testers (Epic Games, Slack) published production-eval quotes; independent reviewer sites (developersdigest, agentbreaking, pandromeda) confirm specs and pricing and cross-check figures. GPT-6 Sol comparison numbers (e.g. GDPval 1487) were taken from OpenAI's publications, not run by Anthropic — treat that column as provider-sourced.