Anthropic's Claude Fable 5.2 Beats GPT-6 Cyber in Multi-Stage Cyberattack Benchmark
What happened
On November 10, 2026, Anthropic released two new AI models, Claude Fable 5.2 and Claude Mythos 5.2. These models demonstrated improvements in coding, reasoning, and vision tasks. Claude Fable 5.2 outperformed OpenAI's GPT-6 Cyber on the Irregular CyScenarioBench, a benchmark for multi-stage cyber operations, by completing a higher percentage of challenges. Anthropic also updated its Claude Security product with new features for code scanning and vulnerability tracking.
Key details
- Claude Fable 5.2 completed 76.3% of CyScenarioBench challenges, compared to 61% for GPT-6 Cyber.
- Claude Fable 5.2 scored 89.6% on Anthropic's CVE Rediscovery benchmark, which included 1,829 open-source advisories from October 2026.
- Claude Mythos 5.2 is described as Anthropic's most capable model for cybersecurity engineering.
- Claude Fable 5.2 is available on Amazon Web Services, Google Cloud, and Microsoft Azure.
- The updated Claude Security now offers continuous code scanning and automatic fix tracking.
Why it matters
Security teams and organizations using AI for cybersecurity now have access to models that can identify and address vulnerabilities more effectively. Anthropic's improvements mean customers can detect more issues in their codebases and receive prioritized, actionable information to fix them. The higher benchmark scores indicate a concrete increase in the ability to automate and support complex cyber defense tasks.