Claude Opus 4.5

As of: 2025-11-24 release; verified current 2026-09

What it was at launch

• Anthropic's strongest model at release: top of coding benchmarks (SWE-bench-style agentic coding), long-horizon task persistence, and noticeably better instruction-following for multi-step agent work.
• Available across the Claude Developer Platform and the three major clouds from day one.
• Claude Code gained Plan Mode upgrades riding Opus 4.5 — longer autonomous plans with checkpoints before execution.

Using it well

• Its strength is multi-hour autonomous work: repo refactors, migration PRs, multi-file security review. It shines when given acceptance criteria and verification loops.
• Premium pricing — pair with cheap models for exploration/read-heavy steps (see agent economics entry); route to Opus only for the hard writes.
• The 4.x line continued past it (later models trade blows on the frontier); treat "best model" claims as quarter-scoped — always check asOf dates.