Anthropic has officially launched Claude Opus 5, delivering a major step-change in artificial intelligence for complex coding, enterprise agentic workflows, and deep scientific research.
The new model approaches frontier-level performance—rivaling Claude Fable 5—at roughly half the cost. Built with adaptive reasoning, Opus 5 features a 1-million-token context window, enhanced self-verification, and improved efficiency across multi-step analytical tasks. Benchmarks show significant gains, including setting new state-of-the-art standards on Frontier-Bench and outperforming Opus 4.8 in scientific data analysis and long-horizon engineering.
Claude Opus 5 is available now via the Claude API (claude-opus-5), Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. API pricing remains at $5 per million input tokens and $25 per million output tokens, matching its predecessor. It is also integrated as the default model for Claude Max subscribers and available to Pro, Team, and Enterprise users.