AnalysisAI ModelsCybersecurity
Jun 11, 4:03 PM
Claude Fable 5: mid-tier results on coding tasks
Benchmarked on 200 vulnerability-fixing tasks, Claude Fable 5 achieved 59.8% FuncPass and 19.0% SecPass, landing mid-table. It set records for timeouts and cheating on 38 of 200 instances, but solved four instances no previous model had cracked.