Yubo Wang et al.

Massive Multitask Language Understanding Professional

An enhanced version of MMLU with 10 answer choices instead of 4, featuring more reasoning-focused questions that better differentiate frontier models.

In index
Knowledge
Source ↗

Models46

Top result89.6%Qwen3.7 Max

Top-3 spread1.1 pts

Year2024

Result and price

Results

46 results

#ModelResult

Qwen3.7 Max89.6%

Claude Opus 4.589.5%

Qwen3.7 Plus88.5%

Qwen3.6 Plus88.5%

Qwen3.5 397B87.8%

DeepSeek V4 Pro 081387.5%

07MAKimi K2.5 (Reasoning)87.1%

07MAKimi K2.587.1%

Nemotron 3 Ultra86.8%

Qwen3.5-122B-A10B86.7%

11UPSolar Pro 486.3%

DeepSeek V4 Flash 073186.2%

12UPSolar Open 286.2%

Qwen3.6-27B86.2%

Qwen3.5-27B86.1%

GLM-585.7%

Qwen3.5-35B-A3B85.3%

Qwen3.6-35B-A3B85.2%

Gemma 4 31B85.2%

MAI-Thinking-185.0%

MiMo-V2-Flash84.9%

GLM-4.784.3%

23LAK-EXAONE 2.083.5%

Qwen3 235B 250783.0%

Gemma 4 26B A4B82.6%

Claude Opus 4.682.0%

27LAExaone 4.0 32B81.8%

Nemotron 3.5 Lightning 30B A3B NVFP481.6%

29MELongCat-Flash-Lite-Sparse79.2%

Claude Sonnet 4.679.2%

31IBGranite 4.2 30B77.6%

Nemotron 3 Nano Omni 30B A3B77.3%

Gemma 4 12B77.2%

34CECeleris-175.9%

DeepSeek V375.9%

36ZYZAYA1-8B74.2%

37IBGranite 4.2 8B74.0%

38OPMiniCPM5-2B70.8%

Gemma 4 E4B69.4%

40ZYZAYA1-74B-Preview68.1%

41IBGranite 4.2 3B67.8%

Gemma 4 E2B60.0%

43SPSoofi S 30B-A3B51.4%

44OPMiniCPM5-1B48.9%

45LILFM2.5-230M20.3%

46LILFM2.5-VL-450M19.3%