Introducing Claude Opus 5
Anthropic News · 2026-07-24
Anthropic launched Claude Opus 5, a model that surpasses all competitors on Frontier-Bench and ARC-AGI 3 while costing the same as its predecessor Opus 4.8, and which Anthropic designates as its most aligned and least deceptive model to date.
Appears in
Extraction
Topics: claude-opus-5anthropic-modelsai-benchmarksmodel-alignmentagentic-ai
Claims
- Claude Opus 5 surpasses all other models on Frontier-Bench v0.1 and more than doubles Opus 4.8's performance at equal or lower cost per task.
- On ARC-AGI 3, Opus 5 scores three times higher than the next-best model, and on Zapier AutomationBench its pass rate is approximately 1.5 times the next-best model at the same cost.
- Anthropic's behavioral audits designate Opus 5 as the company's most aligned model yet, exhibiting lower rates of deceptive behavior and greater resistance to being manipulated into misuse than Fable 5, Sonnet 5, or Opus 4.8.
- Opus 5 does not advance the frontier in dual-use biology or offensive cybersecurity capabilities and remains behind Mythos 5 on exploitation of discovered vulnerabilities.
- Opus 5 is priced identically to Opus 4.8 at $5 per million input tokens and $25 per million output tokens, and a Fast mode runs at approximately 2.5 times default speed at twice the base price.
Key quotes
Opus 5 is our most aligned model to date... It adheres to Claude's Constitution better than Opus 4.8, Sonnet 5, or Fable 5; exhibits the lowest rates of deceptive behavior; and is the least susceptible to being tricked into misuse.
Opus 5 responded by writing its own computer vision pipeline to pull the geometry from the raw pixels, then reconstructed the full machine part. It succeeded in doing so repeatedly; no competing model with the same setup could solve it after five attempts.
On ARC-AGI 3, an evaluation where the model has to solve novel problems, Opus 5's score is three times as high as the next-best model.