thehackernews.com
Anthropic’s Claude Fable 5.2 Beats GPT-6 Cyber in Multi-Stage Cyberattack Benchmark
Anthropic unveiled Claude Fable 5.2 and Claude Mythos 5.2, claiming gains in coding, reasoning, and vision. The models reportedly aided engineers in software migrations, device design, black‑hole physics, robotics, and a rapid Fermat’s Last Theorem proof. On Irregular’s CyScenarioBench, Fable 5.2 achieved 76.3% success, surpassing its predecessor and OpenAI’s GPT‑6 Cyber. It also scored 89.6% on the CVE Rediscovery benchmark, flagging more vulnerabilities than earlier models. The releases are available on AWS, Google Cloud, and Azure.