EntropyMath Leaderboard

Seed

10 problems 10 models

K-LLM Project · 독파모 / EntropyMath 평가 차수 안내

K-LLM Project는 대한민국 과학기술정보통신부의 「독자 AI 파운데이션 모델 프로젝트」(통칭 독파모)을 뜻합니다.

Round 1·2·3은 EntropyMath의 1차·2차·3차 모델 평가를 뜻합니다. 과기정통부의 공식 단계평가 번호나 통과 여부를 뜻하지 않습니다.

프로젝트 공식 안내 공식 1차 단계평가 공식 2차 단계평가

Jump to leaderboard

Leaderboard

Benchmarks ↑
ModelAccPass@30123456789
API / Others
GPT-5.2 (high)
86.790.03/33/33/32/33/33/33/33/30/33/3
Gemini-3-Pro-Preview
86.790.03/33/33/33/33/33/32/33/30/33/3
GLM-4.5-Air
40.060.01/33/33/30/30/32/30/32/30/31/3
K-LLM Project · 2차 평가 (Round 2)
K-EXAONE-236B-A23B
K-EXAONE-236B-A23B
50.060.00/33/33/30/30/33/31/33/30/32/3
Solar-Open-100B
Solar-Open-100B
36.750.00/32/33/30/30/32/30/33/30/31/3
K-LLM Project · 1차 평가 (Round 1)
Solar-Pro 2
Solar-Pro 2
60.070.01/33/33/32/30/33/30/33/30/33/3
HCX-007
HCX-007
30.040.00/30/33/30/30/33/30/32/30/31/3
EXAONE-4.0-32B
EXAONE-4.0-32B
26.740.00/32/33/30/30/32/30/31/30/30/3
Local - KR
Kanana-2-30B-Thinking
Kanana-2-30B-Thinking
53.360.00/33/33/30/30/33/31/33/30/33/3
Kanana-2-30B-Thinking-2601
Kanana-2-30B-Thinking-2601
50.060.00/33/33/31/30/33/30/32/30/33/3
Scroll for problems · Click a cell for the solution
All correctPartialIncorrect

Model Accuracy vs Pass@3

100%
75%
50%
25%
0%
GPT-5.2 (high)
Gemini-3-Pro-Preview
Solar-Pro 2
Solar-Pro 2
Kanana-2-30B-Thinking
Kanana-2-30B-Thinking
K-EXAONE-236B-A23B
K-EXAONE-236B-A23B
Kanana-2-30B-Thinking-2601
Kanana-2-30B-Thinking-2601
GLM-4.5-Air
Solar-Open-100B
Solar-Open-100B
HCX-007
HCX-007
EXAONE-4.0-32B
EXAONE-4.0-32B
Accuracy
Pass@3

Avg Token Usage (Per Problem)

25.2K
18.9K
12.6K
6.3K
0
K-EXAONE-236B-A23B
K-EXAONE-236B-A23B
Solar-Open-100B
Solar-Open-100B
Gemini-3-Pro-Preview
Kanana-2-30B-Thinking-2601
Kanana-2-30B-Thinking-2601
Kanana-2-30B-Thinking
Kanana-2-30B-Thinking
Solar-Pro 2
Solar-Pro 2
GPT-5.2 (high)
GLM-4.5-Air
HCX-007
HCX-007
EXAONE-4.0-32B
EXAONE-4.0-32B
Avg Tokens / Problem
About this benchmark

Curated seed problems that form the foundation of the EntropyMath benchmark.

EntropyMath is an evolutionary multi-agent system and benchmark that generates high-entropy math problems designed to systematically break current LLMs.

Results are reported using Pass@3 metrics to account for generation variance. Detailed execution traces are available for transparency.

Performance Legend

Mastery (100%)
3/3
Strong (66%)
2/3
Weak (33%)
1/3
Fail (0%)
0/3