Claude Opus 5

Claude Opus the listed price is positioned as a thoughtful, proactive model for demanding everyday work, particularly software engineering and knowledge-based tasks. Anthropic says it approaches the frontier intelligence of Claude Fable the listed price at half the price, while improving substantially over Opus 4.8 at the same stated cost. Customers can adjust effort to balance capability against token use, speed, and expense, making the model suitable for workflows that need strong reasoning without automatically running at maximum effort.

The strongest evidence centers on practical coding performance: Anthropic reports state-of-the-art results on Frontier-Bench and GDPval-AA, with more than double Opus 4.8’s performance at a lower cost per task on Frontier-Bench v0.1. It also comes within the listed price% of Claude Fable the listed price’s peak CursorBench 3.2 score at maximum effort while costing half as much per task. Independent code-review testing found cleaner actionable comments but noisier overall review output, so teams may benefit from assigning it a focused role with human or automated quality checks.