Claude Opus 5 is out today and outperforms Claude Fable 5 on most benchmarks while costing about half as much for businesses to run. On Frontier-Bench v0.1, Opus 5 scored 43.3% versus Fable 5’s 33.7%, and on GDPval-AA v2 Opus scored 1,861 compared with Fable’s 1,747. These benchmark figures and the roughly 50% pricing advantage for business users are the headline differences between the two models.
Claude Opus 5 demonstrates superior performance over Claude Fable 5 and GPT-5.6 Sol across several benchmarks. In the Frontier-Bench v0.1 test, Opus 5 achieved a score of 43.3%, outperforming Fable 5, which scored 33.7%, and narrowly exceeding GPT-5.6 Sol’s 34.4%. On the ARC-AGI-3 test, Opus 5 scored 30.2%, whereas GPT-5.6 Sol managed only 7.8%, with Fable 5 not undergoing this test. The GDPval-AA v2 results showed Opus 5 leading with a score of 1,861, compared to Fable 5’s 1,747 and GPT-5.6 Sol’s 1,736. Additionally, Opus 5 achieved a remarkable 100% completion rate on the Zapier AutomationBench’s churn-prevention sequence, a feat previous models could not accomplish.
Claude Fable 5 launched on June 9. Three days after the June 9 launch, the model was pulled globally due to an emergency export control order. The model returned to availability on June 30. When it returned, access to Fable 5 shifted to a credits-only usage model. This sequence—launch, global pull three days later, and return on June 30 with credits-only access—outlines the timeline and operational changes for Claude Fable 5 in the weeks following its release.
Fabian Hedin of Anthropic said Opus 5 is not only better on the company’s hardest agentic coding tasks—showing a 22% improvement over Opus 4.7—but is also steadier, with far less variance from run to run.
Ultima Genomics described the model as behaving more like a careful scientist than previous models, reaching for appropriate statistical tests to rule out confounders, cross-checking results by independent methods, and staying on track through long multi-step analyses.
A reported demonstration said Opus 5 took a raw account-health workbook and ran a full churn-prevention sequence end to end—flagging at-risk accounts, alerting the right owner, and summarizing for retention operations—where previous models did not pass and Opus 5 hit a 100% completion rate.
Claude Opus 5 occupies a strong market position based on its measured outperformance of Claude Fable 5 across multiple benchmark tests and its competitive pricing advantage for business usage, reflecting both higher evaluated capability and reduced operating cost for companies deploying the model. These attributes position Opus 5 as a reliable, cost-efficient option for businesses seeking high-performance AI and indicate a clear competitive choice for organizations weighing capability against operational expense.
