Anthropic: Claude Sonnet 5.5

Claude Sonnet 5.5 is Anthropic's second Claude 5.5 family model, positioned as a faster, lower-cost complement to Claude Opus 5.5 for well-scoped everyday work, software engineering, and polished knowledge-work deliverables such as documents, slides, and spreadsheets. Released on September 28, 2026, it accepts text and image inputs, supports tool use and adaptive thinking, and offers a one-million-token context window with up to 128K output tokens on the synchronous Messages API. Anthropic reports it runs more than 30% faster than Claude Sonnet 5 while often using far fewer tokens at the same list pricing ($2 per million input tokens and $10 per million output tokens).

On Anthropic's published evaluations (Claude Sonnet 5.5 System Card and launch materials, typically adaptive thinking at max effort), Sonnet 5.5 is strongest on agentic software engineering and professional automation. Reported highlights include 81.3% on SWE-Bench Pro, 90.3% on SWE-Bench Multilingual, 71.0% on DeepSWE v1.1, 64.5% on Humanity's Last Exam with tools, 77.8% on Toolathlon-Verified Pass@1, and 44.7% on Zapier's AutomationBench. It also scores 92.1% average accuracy on Global MMLU across 42 languages and an Elo of 1844 on GDPval-AA v2.1 for economically valuable knowledge-work tasks.

Sonnet 5.5 ships with expanded alignment and safety measures relative to earlier Sonnet models, including cyber safeguards comparable to Opus-tier deployments, biology safeguards aligned with Sonnet 5, and classifiers aimed at distillation and reasoning extraction. It is available API-only via the Claude Platform and major cloud partners under the model ID claude-sonnet-5-5, with zero data retention on supported deployments. Anthropic notes that while Sonnet 5.5 reaches near-Opus performance on several benchmarks, Opus 5.5 remains preferable for complex, open-ended work requiring sustained judgment.