What's new in Claude Opus 5 (platform.claude.com)

Claude Opus 5 has been launched as a significant upgrade over its predecessor, Claude Opus 4.8, featuring notable enhancements in deep reasoning, agentic tasks, and test-time compute scaling. With thinking now enabled by default, users can expect the model to autonomously determine its reasoning depth, making it easier to manage complex tasks without constant oversight. Furthermore, the API has introduced features such as mid-conversation tool changes and a new "default" mode for fallback parameters, streamlining interactions and improving user experience in managing tool capabilities during sessions. This iteration marks a transformative shift for the AI/ML community, as Claude Opus 5 delivers frontier-level intelligence at a competitive cost—$5 per million input tokens and $25 per million output tokens—while also enhancing performance across various effort levels. Its ability to generate higher quality outputs with reduced token usage is particularly striking, allowing for efficient handling of low and medium effort tasks. This improvement, coupled with behavioral changes that promote self-verification and better delegation in multi-agent frameworks, positions Claude Opus 5 as a powerful tool for coding and complex enterprise applications, thus setting a new standard in the landscape of AI models.