Anthropic's Claude Opus 5: Everything New, Explained

Anthropic announced the launch of Claude Opus 5 on Monday, describing it as a step-change improvement over its predecessor with major gains in deep reasoning, agentic coding, and long-horizon tasks. The company said the model delivers frontier intelligence at half the cost of its most capable widely available model, Claude Fable 5, while maintaining the same pricing as Opus 4.8 at $5 per million input tokens and $25 per million output tokens. Opus 5 introduces a default extended reasoning mode that lets the model decide when and how deeply to think on each turn, a shift from Opus 4.8 where reasoning had to be manually enabled. Anthropic said this improves answers out of the box but requires developers to revisit output-token limits since thinking and response text now share the same budget. The model converts extra effort into better results more reliably across a new ladder of reasoning tiers, though disabling thinking is only supported at high effort or below. The release highlights stronger agentic coding that completes multi-file features without stubs, improved code review with fewer false positives, and better vision capabilities for charts and documents. Opus 5 handles a 1-million-token context window as standard and writes longer responses by default while narrating progress during agentic sessions. Anthropic advised removing previous verification instructions as the model now verifies its own work automatically. The model is available immediately on the Claude API, AWS, Google Cloud, and Microsoft Foundry.