Introducing Claude Sonnet 5

Claude Sonnet 5 is designed to be the most agentic Sonnet model to date, capable of making plans, using tools like browsers and terminals, and operating autonomously. It narrows the performance gap with higher-tier Opus-class models, providing substantial improvements over its predecessor, Sonnet 4.6, particularly in reasoning, coding, and knowledge work.

The model has undergone safety assessments indicating a lower rate of undesirable behaviors compared to previous versions, including better refusal of malicious requests and resistance to prompt injection attacks. However, it shows reduced capability in cybersecurity tasks relative to Opus models, with safeguards enabled by default to mitigate cyber risks.

Claude Sonnet 5 is now available to all users, including free, pro, Max, team, and enterprise plans, with introductory pricing of $2 per million input tokens and $10 per million output tokens through August 31, 2026. It can be accessed via the Claude API and is integrated into the Claude Platform and Claude Code.

Performance evaluations demonstrate that Sonnet 5 offers improved cost efficiency and broader capability coverage compared to previous models, with adjustable effort levels to balance cost and performance. Early user feedback highlights its enhanced ability to complete complex tasks and self-check outputs, marking a notable step forward in agentic AI development.

This release is likely to influence AI development by providing a more capable and safer agentic model at a competitive price point, potentially accelerating adoption in enterprise and developer communities. The safety improvements and tool use capabilities may also impact how AI safety and alignment are approached in future models.