Frontier Models' Cyber Capabilities Lead General Ability by About a Year
Model capabilities are highly uneven: math and computer science skills far outpace other domains, forming a non-convex 'Shoggoth' frontier. Cyber-ECI analysis shows leading internal models from Anthropic and OpenAI score 175-180 on cyber tasks, about a year ahead of their general capability.
2026-08-27 ~ 2026-08-27 · 2 related posts
- Model Capability Distribution: Uneven Gains in Math and CS — scaling01 · 2026-08-27
- Cyber-ECI reveals frontier models lead general capabilities by ~1 year — scaling01 · 2026-08-27