Claude Sonnet 4.5 is Anthropic's mid-to-large segment model, released in late 2025. The model claims industry leadership on real-world coding benchmarks like SWE-bench Verified, OSWorld, and Terminal-bench, and stands out for sustaining autonomous agent tasks for 30+ hours.
Architecturally a decoder-only Transformer in the Claude family, it ships with an "extended thinking" mode that can deliberate for minutes before answering complex problems. Cursor, Claude Code, and Devin all default to Sonnet 4.5.
- 200K-token context window (extensible to 1M).
- Vision and PDF input supported.
- The most widely deployed production model thanks to its speed/quality balance.
External Links