Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

MMMLU leaderboard — the family board.

3 TRACKED VARIANTS · 1 FIRST-CLASS BOARD

One family, 3 tracked prints. Scoring configs and splits of the base board render as sections of this page, each under its own honest label.

● LIVE DATA80 MODELS ON THIS BOARDLEADER CLAUDE MYTHOS PREVIEW AT 92.7

The board — MMMLUTOP 80 OF 80 MEASURED · BEST SCORE PER MODEL, DEFAULT VARIANT

#ModelVendorScoreVLast seen
01Claude Mythos PreviewAnthropic92.72026-08-23
02Gemini 3.1 ProGoogle92.62026-08-23
03Gemini 3 Flash PreviewGoogle91.82026-08-23
04Gemini 3 ProGoogle91.82026-08-23
05Claude Opus 4.7Anthropic91.52026-08-23
06Claude Opus 4.6Anthropic91.12026-08-23
07GPT-5.1OpenAI912026-07-29
08Claude Opus 4.5Anthropic90.82026-08-23
09Qwen3.7 MaxAlibaba90.32026-08-23
10GPT-5.2OpenAI89.62026-08-23
11Claude Opus 4.1Anthropic89.52026-08-23
12Qwen3.6 PlusAlibaba89.52026-08-23
13Claude Sonnet 4.6Anthropic89.32026-08-23
14Claude Sonnet 4.5Anthropic89.12026-08-23
15Qwen3.7 Plus PreviewAlibaba892026-08-23
16Gemini 3.1 Flash Lite PreviewGoogle88.92026-08-23
17Claude Opus 4Anthropic88.82026-08-23
18Qwen3.5 397B A17BAlibaba88.52026-08-23
19Gemma 4 31BGoogle88.42026-10-08
20o1OpenAI87.72026-08-23
21GPT-4.1OpenAI87.32026-08-23
22Qwen3 235B A22BAlibaba86.72026-08-23
23Qwen3.5 122B A10BAlibaba86.72026-08-23
24Claude Sonnet 4Anthropic86.52026-08-23
25Gemma 4 26B A4BGoogle86.32026-10-08
26Claude 3.7 SonnetAnthropic86.12026-08-23
27Qwen3.5 27BAlibaba85.92026-08-23
28Mistral Large 3Mistral85.52026-08-23
29Mistral Large 3 675B Base 2512Mistral85.52026-08-23
30Qwen3.5 35B A3BAlibaba85.22026-08-23
31GPT-4.5 PreviewOpenAI85.12026-08-23
32Qwen3 Max (Reasoning)Alibaba84.42026-09-09
33Gemma 4 12BGoogle83.42026-10-08
34Claude Haiku 4.5Anthropic832026-08-23
35GPT-4oOpenAI81.42026-08-23
36Qwen3.5 9BAlibaba81.22026-08-23
37Hy3-preview-BaseTencent80.22026-06-05
38DeepSeek-V3-BaseDeepSeek79.52026-06-05
39DeepSeek-V3DeepSeek79.42026-07-06
40GLM-4.5-BaseZ.ai79.32026-06-05
41Qwen3 32BAlibaba792026-06-15
42Qwen3 30B A3B 2507 ThinkingAlibaba792026-06-15
43GPT-4.1 miniOpenAI78.52026-08-23
44Seed Oss 36B InstructByteDance78.42026-06-15
45Kimi K2 BaseMoonshot77.62026-06-05
46Gemma 4 E4BGoogle76.62026-10-08
47Qwen3.5 4BAlibaba76.12026-08-23
48Qwen2.5 Instruct 72BAlibaba74.82026-07-06
49Llama 3.1 Instruct 405BMeta73.82026-07-06
50Granite 4.1 30BIBM73.72026-06-15
51Granite 4.1 30B BaseIBM73.42026-06-15
52Gemma 3 27BGoogle70.72026-10-08
53Phi 3.5 MoE InstructMicrosoft69.92026-08-23
54Gemma 4 E2BGoogle67.42026-10-08
55GPT-4.1 nanoOpenAI66.92026-08-23
56Qwen2.5 7B BaseAlibaba65.52026-05-03
57Granite 4.1 8BIBM64.82026-06-15
58Granite 4.1 8B BaseIBM64.72026-06-15
59DeepSeek-V2DeepSeek642026-07-06
60Qwen3.5 2BAlibaba63.12026-08-23
61Qwen3 4B 2507 InstructAlibaba60.72026-10-07
62Qwen2.5 3B BaseAlibaba59.62026-05-03
63Granite 4.1 3BIBM57.62026-06-15
64Granite 4.1 3B BaseIBM56.62026-06-15
65Granite 4.0 H TinyIBM56.12026-10-07
66Phi 3.5 Mini InstructMicrosoft55.42026-08-23
67LFM2-2.6BLiquid AI55.42026-10-07
68LFM2-8B-A1B (Non-Reasoning)Liquid AI55.32026-10-07
69Qwen3 4B Base DenseAlibaba50.22026-05-03
70Gemma 3 4BGoogle50.12026-10-07
71Llama 3.2 3B InstructMeta47.92026-10-07
72LFM2-1.2BLiquid AI46.72026-10-07
73Qwen3 1.7BAlibaba46.52026-10-07
74Qwen3.5 0.8BAlibaba44.32026-08-23
75LFM2-700MLiquid AI43.32026-10-07
76Llama 3.2 Instruct 1BMeta38.12026-10-07
77LFM2-350MLiquid AI38.02026-10-07
78Gemma 3 1B InstructGoogle34.42026-10-07
79Qwen3 0.6BAlibaba30.82026-10-07
80Gemma 3 1B Base DenseGoogle26.62026-05-03
◆ IV — INDEPENDENTLY VERIFIED · ◆ AA — AGGREGATOR · ◆ VA — VENDOR · ◇ SA — SOURCE-ATTRIBUTED

Variants of this boardSCORING CONFIGS & SPLITS — NOT CITED STANDALONE, SO THEY LIVE HERE

MMMLU by language

PER-LANGUAGE1 single-language boards, one table
LanguageLeaderScoreModels
Non English AccDeepSeek-V379.44

Other tracked prints

THIN BOARDSunder 5 published models each — listed, never hidden
PrintLeaderScoreModels
MMMLU (EM)DeepSeek-V4-Pro-Base90.33
THE FAMILY BOARD · EVERY PRINT LABELED · EVERY SCORE SOURCE-ATTRIBUTED ON ITS OWN LIVE BOARDAll leaderboardsThe frontier boardModels by vendor