Claude Sonnet 4.5

Anthropic's state-of-the-art model: the world's best for agentic workflows, coding, and autonomous computer use.

Claude Sonnet 4.5 is the new standard for enterprise AI: a powerful, mid-sized model for complex, real-world applications. It excels as the best coding model globally, leading on the SWE-bench Verified evaluation and achieving a 0% error rate on Anthropic's internal code editing benchmark. The model is engineered for advanced agentic capabilities, demonstrating superior instruction following, parallel tool execution, and maintaining focus for over 30 hours on autonomous coding tasks. Developers access it via the `claude-sonnet-4-5` API, and it's available across major cloud platforms, including Amazon Bedrock and Google Cloud's Vertex AI. This release also marks Anthropic's most aligned frontier model, significantly reducing concerning behaviors like sycophancy and fortifying defenses against prompt injection attacks.

https://claude.ai