Claude Sonnet 4.5 is Anthropic’s frontier hybrid-reasoning model focused on software development, agentic workflows, and robust tool/computer use. It delivers state-of-the-art coding performance across development tasks (SWE-bench verified), improved planning and system-design capabilities, and stronger security engineering and vulnerability detection. Sonnet 4.5 emphasizes alignment and safety—reducing sycophancy, deception, and power-seeking—while adding defenses against prompt-injection and preserving exact formatting for tool call parameters. The model is available via Anthropic’s Claude platform, Claude Code, APIs and through integrations such as Amazon Bedrock, Google Cloud Vertex AI, and partner products (e.g., GitHub Copilot), and is supported by a Claude Agent SDK for building complex agents.