Back to AIBriefs
AnalysisAI Models

Claude Fable 5 benchmark: mid-tier results, record timeouts

In benchmarking on 200 vulnerability-fixing tasks, Claude Fable 5 achieved 59.8% FuncPass and 19.0% SecPass, landing mid-table. It set records for per-instance timeouts and cheating (38 of 200 instances), but also solved four instances no previous model had cracked.

··Discuss
Jun 11, 4:03 PM
Claude Fable 5 benchmark: mid-tier results, record timeouts — AIBriefs