Epoch AI

FrontierMath-2025-02-28-Public

Evaluated by Epoch AI on their own harness.

Not in index
Math
Best in math
Source ↗

Models64

Top result100.0%Claude Fable 5

Top-3 spread10.0 pts

Year—

Result and price

48 of 64 results

Results

64 results

#ModelResult

01Claude Fable 5max effort100.0%

02Claude Opus 4.7xhigh effort90.0% ±10.0

02Kimi K2.690.0% ±10.0

02Claude Opus 4.6max effort90.0% ±10.0

05Gemini 3.1 Pro88.9% ±11.1

06GPT-5.4high effort80.0% ±13.3

06Gemini 3.5 Flashhigh effort80.0% ±13.3

06Gemini 3 Pro80.0% ±13.3

06Claude Sonnet 4.680.0% ±13.3

06GLM-5.180.0% ±13.3

11GPT-5.2high effort60.0% ±16.3

11GPT-5high effort60.0% ±16.3

11Gemini 3 Flash60.0% ±16.3

11GPT-5.4 nanohigh effort60.0% ±16.3

11GPT-5 Prohigh effort60.0% ±16.3

16Qwen 3.6 Max (preview)50.0% ±16.7

16Qwen3.6 Plus50.0% ±16.7

16GPT-5.4 minihigh effort50.0% ±16.7

16Qwen 3.5 Plus (hosted 397B-A17B)50.0% ±16.7

20o3-minihigh effort40.0% ±16.3

21Gemini 2.5 Pro (Jun 2025)35.0% ±16.3

22Claude Opus 4.530.0% ±16.3

22GPT-5 minihigh effort30.0% ±15.3

22Kimi K2 Thinking30.0% ±16.3

22o4-minihigh effort30.0% ±15.3

26GPT-5 nanohigh effort20.0% ±13.3

26Qwen3-235B-A22B (Jul 2025)20.0% ±13.3

26Qwen3.6 Flash20.0% ±13.3

29o3high effort10.0% ±10.0

29GPT-4.1 mini10.0% ±10.0

29Qwen 3.5 Flash (hosted 35B-A3B)10.0% ±10.0

32GLM-52.1% ±2.1

33Claude Sonnet 4.50.0%

33Claude Opus 4.10.0%

33o1high effort0.0%

33Grok 3 Minihigh effort0.0%

33Claude Opus 40.0%

33Claude Haiku 4.50.0%

33Claude Sonnet 40.0%

33Grok 3 [Beta]0.0%

33Qwen3 235B-A22B0.0%

33Claude 3.7 Sonnet0.0%

33Llama 4 Maverick0.0%

33o1-minihigh effort0.0%

33Claude 3.5 Sonnet0.0%

33DeepSeek V30.0%

33Gemini 2.0 Flash0.0%

33Grok 20.0%

33GPT-4o0.0%

33Mistral Large 20.0% ±10.0

33Mistral Medium 30.0%

33Llama 4 Scout0.0%

33GPT-4.1 nano0.0%

33Claude 3.5 Haiku0.0%

33Gemini 1.5 Flash0.0%

33Claude 3.5 Sonnet (October 2024)0.0%

33DeepSeek-V3 (Mar 2025)0.0%

33Gemini 2.0 Pro0.0%

33Magistral Small 1.00.0%

33Qwen Plus0.0%

33Qwen2.5-Max0.0%