Evaluation & BenchmarksEpoch AI (Elliot Glazer et al.)Epoch AI

Epoch AI releases FrontierMath

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

FrontierMath released — hundreds of unpublished math problems that take researchers hours to days to solve. At release, the SOTA solve rate was under 2%

Why it matters

A holdout design that cannot be contaminated

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2411.04872

Added in Week 35, 2026 · See that week

On the map — ⑯ Evaluation line

  1. Previous stopGSM1k
  2. This recordFrontierMath
  3. Next stopTechCrunch reports criticism of Epoch AI for not disclosing OpenAI funding
See it on the map

Same week

The Information reports slowing gains for OpenAI's OrionPressCONFIRMEDPressMicrosoft Research presents Magentic-OnePaperTanishq Kumar et al. publish "Scaling Laws for Precision"Paper

Nearby on the same axis

ARC Prize releases o3's ARC-AGI resultsOfficialTechCrunch reports criticism of Epoch AI for not disclosing OpenAI fundingPressEpoch AI posts a clarification on how FrontierMath was made and is usedOfficialOpenAI releases SWE-bench VerifiedOfficial