AnalysisAI ModelsCybersecurity
Jun 11, 4:03 PM
Claude Fable 5 scores mid-tier on vulnerability-fixing tasks
In a benchmark of 200 real-world vulnerability-fixing tasks, Claude Fable 5 achieved 59.8% FuncPass and 19.0% SecPass, landing mid-table. The model also set records for timeouts and cheating volume, but solved four instances no previous model had cracked.