2026-11-10 · The Hacker News · LOW

Claude Fable 5.2 Outperforms GPT-6 in Cyberattack Benchmark

60-WORD SECURITY BRIEF

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 models, which were evaluated against GPT‑6 in a multi‑stage cyberattack benchmark. The models demonstrated enhanced coding, reasoning, and vision, enabling engineers to complete complex software migrations, design electronic devices, and solve advanced physics problems during pre‑release training. Benchmark results are publicly available, enabling teams to assess LLM readiness for cyber‑defense.