Claude Sonnet 4.5
Claude Sonnet 4.5 reaches 77.2% on SWE-bench Verified and 61.4% on OSWorld, and Anthropic reports it staying on task for 30+ hours.
Positioned by Anthropic as its best model for agents, coding and computer use, priced as Sonnet 4 ($3/$15). SWE-bench Verified 77.2%, OSWorld 61.4%. Ships with the Claude Agent SDK, Claude Code checkpoints and a VS Code extension, plus API context editing and a memory tool. Released under ASL-3.
- Date
- Monday, 29 September 2025
- Lab
- Anthropic
- Kind
- model
- Access
- closed API
- Price
- $3/$15 per M tokens in/out, same as Sonnet 4
Figures
| Measure | Value | Measured by |
|---|---|---|
| SWE-bench Verified | 77.2% as reported on the launch page | company |
| OSWorld | 61.4% computer-use benchmark | company |
| Sustained autonomous focus | 30+ hours on complex multi-step tasks, company-reported | company |
Deprecated 2026-09-30 with retirement set for 2026-11-30 (API release notes).
Sources
This record was checked against its sources on 6 October 2026. How we check