Anthropic released Claude Sonnet 4.5 on September 29, 2025, as the latest model in the Claude 4 family, featuring enhanced coding performance, support for long-running agent workflows, and more reliable computer-use tasks.
- Scores 77.2% on SWE-bench Verified (82.0% with parallel compute) and 50.0% on Terminal-Bench.
- Achieves 100% on AIME 2025 with Python tools and 61.4% on OSWorld for computer interaction.
- Supports runs exceeding 30 hours, reducing shortcut behavior by 65% compared to Sonnet 3.7.
- Offers a 200K token context window (with 1M option) and outputs up to 64K tokens.
- Priced at $3 per million input tokens and $15 per million output tokens, available via API, web, mobile, Amazon Bedrock, and Google Cloud Vertex AI.
The model aims to provide developers with stronger reliability for complex coding and multi-session agent workflows while maintaining competitive pricing against rivals like GPT-5.