Claude Opus 5 is now documented: Anthropic’s latest Opus model (replaces Opus 4.8), with stronger agentic coding and long-horizon reasoning at standard Opus pricing; includes a high-thinking variant, Fast mode (claude-opus-5-fast at $10/…
Changelog
Models
Claude Opus 5 is now documented: Anthropic’s latest Opus model (replaces Opus 4.8), with stronger agentic coding and long-horizon reasoning at standard Opus pricing; includes a high-thinking variant, Fast mode (claude-opus-5-fast at $10/M input and $50/M output), context up to 1M at base rates, and Zero Data Retention compatibility.
Older Opus and Sonnet pages now recommend Opus 5; Fable 5 docs compare against Opus 5 (still ~2× Opus price, with Anthropic data-retention opt-in).
Enterprise
Cyber Verification Program (CVP) support extended to Opus 5 (alongside Opus 4.7 and 4.8).
Agent
Search docs retitled and narrowed to Instant Grep and the Explore subagent (semantic search section removed).
Subagent and evals examples updated to claude-opus-5.