FrontierMath Tier 4 Leaderboard 2026 — Results Across 2 Real AI Models

FrontierMath Tier 4 leaderboard

FrontierMath Tier 4

2 models tested · Updated 2026-03-05 · Verified sources only

      GPT-5.4 Pro leads at 38.0%
    

OpenAI · Epoch AI Blog · 2026-03-05

Research-grade math problems. Solved 2 previously unsolved problems. Best Tier 4 score to date.

38.0%

Meta · X/@EpochAIResearch · 2026-04-08

Independent evaluation by Epoch AI on Tier 4 (research-level math). Behind GPT-5.4 Pro (38%).

15.0%