MMLU-Pro

14개 과목 영역에 걸친 12,000개의 대학원 수준 문항으로 MMLU를 강화한 버전으로, 10개의 답안 선택지와 더 깊은 추론 요구사항을 갖추고 있습니다.

모델: 83 현재, 349 전체
중앙값: 74.2%
범위: 5.5%–85.4%

점수 분포

가장 낮음: 5.5%
중앙값: 74.2%
최고: 85.4%

리더보드

지원 중단된 모델 포함

순위	모델	점수
1	Doubao Seed Code ByteDance Seed	85.4%
2	o3 OpenAI	85.3%
3	Cogito v2.1 (Reasoning) Deep Cogito	84.9%
4	ERNIE 5.0 Thinking Preview Baidu	83.0%
5	Nova 2.0 Pro Preview (Medium) Amazon	83.0%
6	Hermes 4 - Llama-3.1 405B (Reasoning) Nous Research	82.9%
7	Llama 3.1 Nemotron Ultra 253B v1 (Reasoning) NVIDIA	82.5%
8	Qwen3 Next 80B A3B (Reasoning) Alibaba	82.4%
9	INTELLECT-3 (Based on GLM-4.5-Air) Prime Intellect	82.2%
10	Nova 2.0 Pro Preview (Low) Amazon	82.2%
11	Qwen3 Next 80B A3B Instruct Alibaba	81.9%
12	Nova 2.0 Lite (High) Amazon	81.8%
13	EXAONE 4.0 32B (Reasoning) LG AI Research	81.8%
14	Magistral Medium 1.2 Mistral	81.5%
15	Nova 2.0 Lite (Medium) Amazon	81.3%
16	Hermes 4 - Llama-3.1 70B (Reasoning) Nous Research	81.1%
17	Mi:dm K 2.5 Pro Korea Telecom	80.9%
18	Nova 2.0 Omni (Medium) Amazon	80.9%
19	Llama 4 Maverick Meta	80.9%
20	gpt-oss-120b (High) OpenAI	80.8%
21	Nova 2.0 Omni (Low) Amazon	79.8%
22	Motif-2-12.7B-Reasoning Motif Technologies	79.6%
23	NVIDIA Nemotron 3 Nano 30B A3B (Reasoning) NVIDIA	79.4%
24	Ring-flash-2.0 InclusionAI	79.3%
25	Qwen3 Omni 30B A3B (Reasoning) Alibaba	79.2%
26	Apriel-v1.6-15B-Thinker ServiceNow	79.0%
27	Nova 2.0 Lite (Low) Amazon	78.8%
28	K2-V2 (High) Institute of Foundation Models	78.6%
29	HyperCLOVA X SEED Think (32B) Naver	78.5%
30	Celeris-1 Celeris	78.0%
31	ERNIE 4.5 300B A47B Baidu	77.6%
32	gpt-oss-120b (Low) OpenAI	77.5%
33	Nova 2.0 Pro Preview (Non-reasoning) Amazon	77.2%
34	Magistral Small 1.2 Mistral	76.8%
35	EXAONE 4.0 32B (Non-reasoning) LG AI Research	76.8%
36	Olmo 3.1 32B Think Allen Institute for AI	76.3%
37	K2-V2 (Medium) Institute of Foundation Models	76.1%
38	NVIDIA Nemotron Nano 12B v2 VL (Reasoning) NVIDIA	75.9%
39	Llama 4 Scout Meta	75.2%
40	gpt-oss-20b (High) OpenAI	74.8%
41	Nova 2.0 Lite (Non-reasoning) Amazon	74.3%
42	NVIDIA Nemotron Nano 9B V2 (Reasoning) NVIDIA	74.2%
43	NVIDIA Nemotron Nano 9B V2 (Non-reasoning) NVIDIA	73.9%
44	Nova Premier Amazon	73.3%
45	Llama 3.1 Instruct 405B Meta	73.2%
46	Hermes 4 - Llama-3.1 405B (Non-reasoning) Nous Research	72.9%
47	Falcon-H1R-7B TII UAE	72.5%
48	Qwen3 Omni 30B A3B Instruct Alibaba	72.5%
49	Nova 2.0 Omni (Non-reasoning) Amazon	71.9%
50	gpt-oss-20b (Low) OpenAI	71.8%

1
Doubao Seed Code
ByteDance Seed
85.4%
2
o3

데이터 출처: Artificial Analysis

Command Palette

MMLU-Pro

점수 분포

리더보드