Benchmark RadarAI BENCHMARK PROFILE
FrontierCode
FrontierCode is Cognition's coding evaluation that tests whether models can pass difficult coding tasks while meeting the standards of high-quality production codebases. The Diamond subset contains the hardest problems.
- Released
- Unknown
- Readiness
- Paper only
- Primary field
- General AI
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.