Reported 76.3% vs GPT-6 Cyber's 61% on multi-stage attack challenges; Epoch Capabilities Index record 174. Chainguard says its Athena system used Mythos to find a dozen mostly pre-auth RCE flaws in six hours; two Xen hypervisor flaws were found with Claude.
Claude Fable 5.2 and Claude Mythos 5.2 are available now:
⚡ Anthropic's new Claude Fable 5.2 beats OpenAI's GPT-6 Cyber in multi-stage cyberattack tests. 76.3% vs. 61% challenge completion. It also scored 89.6% on a benchmark for rediscovering known vulnerabilities. Learn more ↓
Anthropic Rolls Out Mythos 5.2 With Improved Bug Hunting, Fewer Sandbox Escapes
Claude Fable 5.2 sets a new record on the Epoch Capabilities Index (ECI) at 174. The previous high was GPT-6 Astra’s 169. Fable 5.2’s score is above what the pre-Astra trend for reasoning models would have predicted. This is notable given recent discussion of
Over a dozen vulnerabilities found in 6 hours. Most were pre-auth RCE flaws. Chainguard says its Athena system used Anthropic's Mythos to identify the flaws in a heavily tested open-source project. Every finding was reproduced 3 times in a clean VM. The proj
‼️ Two Xen hypervisor flaws found using Anthropic's Claude allow a malicious guest VM to read and write memory belonging to other VMs, including privileged dom0. • Xen 4.6+ affected • Patches for 4.17+, none for 4.6-4.16 • No known workaround • Qubes OS and X
Why this score
Plan rubric with caps (verified sweep, Nov 11)