← Timeline · Benchmark
FrontierMath: frontier models solve under 2%
Epoch AI's research-level maths benchmark shows leading models solving fewer than 2% of problems.
What happened and why it matters
FrontierMath consists of unpublished problems written by research mathematicians, many requiring hours or days for experts. The near-zero scores were a sobering counterpoint to saturated school-level benchmarks.
Sources
Hover the chart to read values · click a marker to open its articles