benchgap
Knowledge & reasoning

MMLU-Pro leaderboard

As of 2026-11-21, the highest measured score on MMLU-Pro is 89.6% by Qwen3.7 Max. 237 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Qwen3.7 Max89.6%measured
2Claude Opus 4.589.5%measured
3Pareto 26.10 Preview88.8%estimated ± 5.4 pp, medium confidence
4Qwen3.6 Plus88.5%measured
5Qwen3.7 Plus88.5%measured
6Claude Fable 588.5%estimated ± 4.8 pp, medium confidence
7Qwen3.5 397B87.8%measured
8Gemini 3.8 Flash87.8%estimated ± 4.8 pp, medium confidence
9Gemini 3.7 Flash87.7%estimated ± 4.8 pp, medium confidence
10Gemini 3 Pro87.7%estimated ± 4.8 pp, medium confidence
11DeepSeek V4 Pro87.5%measured
12Grok 4.687.3%estimated ± 4.8 pp, medium confidence
13Gemini 3.6 Flash87.3%estimated ± 4.8 pp, high confidence
14Grok 4.587.2%estimated ± 4.8 pp, high confidence
15Claude Opus 4.6 (Adaptive)87.2%estimated ± 4.8 pp, high confidence
16Kimi K2.587.1%measured
17Mercury 387.1%estimated ± 5.4 pp, medium confidence
18Gemini 3 Flash86.9%estimated ± 4.8 pp, high confidence
19Nemotron 3 Ultra86.8%measured
20Muse Spark 1.286.7%estimated ± 4.8 pp, high confidence
21Qwen3.5-122B-A10B86.7%measured
22OrcaCyber-Nano-1.086.6%estimated ± 5.4 pp, medium confidence
23Solar Pro 486.3%measured
24Claude Sonnet 4.5 Thinking86.2%estimated ± 4.8 pp, high confidence
25DeepSeek V4 Flash 073186.2%measured
26Qwen3.6-27B86.2%measured
27Solar Open 286.2%measured
28Claude Opus 4.5 Thinking86.2%estimated ± 4.8 pp, high confidence
29Claude Fable 5.286.2%estimated ± 2.5 pp, low confidence
30Qwen3.5-27B86.1%measured
31o3-pro86.1%estimated ± 5.4 pp, medium confidence
32Claude Opus 5.586.1%estimated ± 2.5 pp, low confidence
33Claude Fable 5.186.0%estimated ± 2.5 pp, low confidence
34Claude Mythos 586.0%estimated ± 2.5 pp, low confidence
35Claude Sonnet 5.586.0%estimated ± 2.5 pp, low confidence
36GLM-5.186.0%estimated ± 4.8 pp, high confidence
37Claude Opus 586.0%estimated ± 2.5 pp, low confidence
38GLM-5.385.9%estimated ± 4.8 pp, high confidence
39Muse Spark 1.185.9%estimated ± 2.5 pp, low confidence
40Qwen4 Max85.9%estimated ± 2.5 pp, low confidence
41Kimi K3.1 Pro85.9%estimated ± 2.5 pp, low confidence
42Sakana Fugu-Ultra85.8%estimated ± 2.5 pp, low confidence
43Claude Opus 4.885.8%estimated ± 2.5 pp, low confidence
44o185.8%estimated ± 4.0 pp, medium confidence
45Pareto 26.985.8%estimated ± 2.5 pp, low confidence
46Sakana Fugu85.8%estimated ± 2.5 pp, low confidence
47Claude Opus 4.785.8%estimated ± 2.5 pp, low confidence
48ERNIE 5.5 Flash Preview85.8%estimated ± 2.5 pp, low confidence
49Claude Haiku 5.585.8%estimated ± 2.5 pp, low confidence
50Gemini 3.1 Pro85.7%estimated ± 2.5 pp, low confidence
51Claude 4.1 Opus85.7%estimated ± 5.4 pp, medium confidence
52GPT-585.7%estimated ± 4.8 pp, high confidence
53Ornith-1.5-397B85.7%estimated ± 2.5 pp, low confidence
54Inkling 2 Small85.7%estimated ± 2.5 pp, low confidence
55GLM-585.7%measured
56Qwen3.8 Max85.7%estimated ± 2.5 pp, low confidence
57Kimi K385.7%estimated ± 2.5 pp, low confidence
58Hy4 preview85.7%estimated ± 2.5 pp, low confidence
59MiniMax M3.1 Flash85.7%estimated ± 2.5 pp, low confidence
60Claude Sonnet 585.7%estimated ± 2.5 pp, low confidence
61GPT-5.5 Pro85.7%estimated ± 2.5 pp, low confidence
62Ling 3.1 Flash85.7%estimated ± 14.4 pp, low confidence
63Muse Spark85.7%estimated ± 2.5 pp, low confidence
64GPT-5.185.7%estimated ± 4.8 pp, high confidence
65GPT-5.4 Pro85.7%estimated ± 2.5 pp, low confidence
66GPT-6.2 Astra85.6%estimated ± 4.4 pp, medium confidence
67GPT-5.585.6%estimated ± 2.5 pp, low confidence
68GLM-5.285.6%estimated ± 2.5 pp, low confidence
69GPT-5.485.6%estimated ± 2.5 pp, medium confidence
70Seed 2.1 Pro85.6%estimated ± 6.5 pp, medium confidence
71Gemini 3.1 Flash-Lite85.6%estimated ± 4.8 pp, high confidence
72GPT-6.2 Sol85.6%estimated ± 4.4 pp, medium confidence
73GPT-6 Astra85.6%estimated ± 4.4 pp, medium confidence
74GPT-6.1 Sol85.5%estimated ± 5.0 pp, low confidence
75Gemini 4 Neon85.5%estimated ± 5.0 pp, low confidence
76Gemini 4 Carbon85.5%estimated ± 5.0 pp, low confidence
77GPT-6 Sol85.5%estimated ± 5.0 pp, low confidence
78Gemini 4 Hydrogen85.5%estimated ± 5.0 pp, low confidence
79Gemini 4 Argon85.5%estimated ± 5.0 pp, medium confidence
80GPT-6.1 Terra85.5%estimated ± 5.0 pp, medium confidence
81Muse Spark 1.385.5%estimated ± 5.0 pp, medium confidence
82Grok 4.785.5%estimated ± 5.0 pp, medium confidence
83Muse Spark Open85.5%estimated ± 5.0 pp, medium confidence
84GPT-5.2-Codex85.5%estimated ± 5.0 pp, medium confidence
85Muse Wonder Full85.5%estimated ± 5.0 pp, medium confidence
86GLM-5.3-Flash85.5%estimated ± 4.8 pp, high confidence
87GPT-5.1-Codex85.5%estimated ± 5.0 pp, medium confidence
88GPT-5.1-Codex-Max85.5%estimated ± 5.0 pp, medium confidence
89GPT-6 Luna85.5%estimated ± 5.0 pp, medium confidence
90Muse Wonder85.5%estimated ± 5.0 pp, medium confidence
91Ling 3.0 Tiny85.5%estimated ± 5.0 pp, medium confidence
92Qwen 3.6 Max (preview)85.5%estimated ± 5.0 pp, medium confidence
93Step 5 Preview85.5%estimated ± 5.0 pp, medium confidence
94MiMo-V2.6-Pro85.5%estimated ± 5.0 pp, medium confidence
95Beam 1.0 501B-A23B85.5%estimated ± 2.5 pp, medium confidence
96Beam 1.0 501B-A23B FP885.5%estimated ± 2.5 pp, medium confidence
97Qwen3.8-Flash-Next85.5%estimated ± 2.5 pp, medium confidence
98GPT-5.6 Sol85.4%estimated ± 4.4 pp, medium confidence
99Beam 1.0 501B-A23B NVFP485.4%estimated ± 2.5 pp, medium confidence
100MiMo-V2.5-Pro85.4%estimated ± 2.5 pp, medium confidence
101Qwen3.8 Max Preview85.3%estimated ± 5.0 pp, medium confidence
102Gemini 3.5 Flash-Lite85.3%estimated ± 4.8 pp, high confidence
103GLM-5V-Turbo85.3%estimated ± 5.0 pp, medium confidence
104Qwen3.5-35B-A3B85.3%measured
105GLM-5-Turbo85.3%estimated ± 5.0 pp, medium confidence
106Nemotron 3.5 Super VL 120B A12B85.3%estimated ± 2.5 pp, medium confidence
107Apodex 1.185.3%estimated ± 5.0 pp, medium confidence
108Apodex 1.1 Mini85.3%estimated ± 5.0 pp, medium confidence
109Grok 4.2085.3%estimated ± 2.5 pp, medium confidence
110Inkling-Small85.3%estimated ± 2.5 pp, medium confidence
111Muse Glimmer 30B85.3%estimated ± 5.0 pp, medium confidence
112Nemotron 3.5 Super VL 120B A12B NVFP485.3%estimated ± 2.5 pp, medium confidence
113GPT-5.6 Terra85.2%estimated ± 4.4 pp, medium confidence
114Qwen3.8-27B85.2%estimated ± 2.5 pp, medium confidence
115Inkling 2 Mini85.2%estimated ± 2.5 pp, medium confidence
116o385.2%estimated ± 4.8 pp, high confidence
117MiMo-V2.6-Flash85.2%estimated ± 5.0 pp, medium confidence
118Gemma 4 31B85.2%measured
119Qwen3.6-35B-A3B85.2%measured
120GPT-5.285.2%estimated ± 4.4 pp, high confidence
121Inkling85.2%estimated ± 2.5 pp, medium confidence
122GPT-5.6 Luna85.2%estimated ± 4.4 pp, high confidence
123Gemini 3.5 Flash85.2%estimated ± 4.4 pp, high confidence
124MiMo-V2-Pro85.2%estimated ± 5.0 pp, medium confidence
125Gemini 2.5 Flash85.1%estimated ± 5.0 pp, medium confidence
126Mistral Large 485.1%estimated ± 5.0 pp, medium confidence
127Nemotron 3.5 Bolt 74B-A9B BF1685.1%estimated ± 2.5 pp, medium confidence
128GPT-5.3 Codex85.1%estimated ± 7.0 pp, medium confidence
129Step 3.7 Flash85.1%estimated ± 5.0 pp, medium confidence
130GPT-5.4 mini85.1%estimated ± 2.5 pp, medium confidence
131Grok 485.0%estimated ± 4.8 pp, high confidence
132Qwen3.8-Omni-Flash85.0%estimated ± 4.4 pp, high confidence
133DeepSeek V4.1 Flash 101785.0%estimated ± 4.4 pp, high confidence
134DeepSeek-R1-052885.0%measured
135MAI-Thinking-185.0%measured
136Kimi K2.685.0%estimated ± 4.4 pp, high confidence
137Hy384.9%estimated ± 7.0 pp, medium confidence
138Kimi K2.7 Code84.9%estimated ± 7.0 pp, medium confidence
139Grok 4.384.9%estimated ± 4.4 pp, high confidence
140DeepSeek V3 032484.9%estimated ± 5.0 pp, medium confidence
141Ornith-1.5-35B-A3B84.9%estimated ± 2.5 pp, medium confidence
142Interfaze Beta84.9%estimated ± 4.4 pp, high confidence
143MiMo-V2-Flash84.9%measured
144Nemotron 3.5 Bolt 74B-A9B84.9%estimated ± 2.5 pp, medium confidence
145DeepSeek V3.284.9%estimated ± 5.0 pp, medium confidence
146GPT-5.4 nano84.8%estimated ± 2.5 pp, medium confidence
147Grok Code Fast 184.8%estimated ± 5.0 pp, medium confidence
148GPT-4.584.7%estimated ± 3.6 pp, medium confidence
149Claude 4 Sonnet84.6%estimated ± 5.0 pp, medium confidence
150Ling 3.0 Flash VL84.6%estimated ± 7.0 pp, medium confidence
151Hy3 Preview84.5%estimated ± 4.4 pp, high confidence
152Qwen3 Max84.5%estimated ± 4.8 pp, high confidence
153Qwen3 235B-A22B (Reasoning)84.5%estimated ± 1.1 pp, high confidence
154Seed 2.1 Turbo84.5%estimated ± 6.5 pp, medium confidence
155Mercury 2.584.4%estimated ± 5.0 pp, medium confidence
156Grok 4.1 Fast (Reasoning)84.4%estimated ± 4.8 pp, high confidence
157MiniMax M384.4%estimated ± 4.8 pp, high confidence
158Ornith-1.5-9B84.4%estimated ± 2.5 pp, medium confidence
159Qwen3.5 Flash84.4%estimated ± 4.8 pp, high confidence
160GLM-4.784.3%measured
161Mistral Small 484.2%estimated ± 5.0 pp, medium confidence
162Ling 3.0 Flash84.2%estimated ± 4.4 pp, high confidence
163Ripple-65B Preview84.2%measured
164Muse Ember84.1%estimated ± 5.0 pp, medium confidence
165Spark-X2.6-4B84.1%estimated ± 2.5 pp, medium confidence
166Ling 3.0 Flash FP884.1%estimated ± 4.4 pp, high confidence
167DeepSeek-R184.0%measured
168Claude Sonnet 4.584.0%estimated ± 4.4 pp, high confidence
169Gemini 2.5 Pro83.9%estimated ± 4.4 pp, high confidence
170Nemotron 3 Super 100B83.7%measured
171DeepSeek V3.183.7%measured
172MiMo-V2.583.7%estimated ± 4.8 pp, high confidence
173Mistral Large 283.6%estimated ± 5.0 pp, medium confidence
174K-EXAONE 2.083.5%measured
175Grok 3 Mini83.4%estimated ± 4.4 pp, high confidence
176Qwen3.5-4B83.3%estimated ± 7.0 pp, medium confidence
177MiMo-V2-Omni83.3%estimated ± 5.0 pp, medium confidence
178Claude 4 Opus Thinking83.3%estimated ± 4.4 pp, high confidence
179GLM-4.683.2%estimated ± 4.8 pp, high confidence
180o1-pro83.1%estimated ± 4.4 pp, high confidence
181Qwen3 235B 250783.0%measured
182Claude 3.7 Sonnet Thinking83.0%estimated ± 4.4 pp, high confidence
183Solar Pro 382.7%estimated ± 5.0 pp, medium confidence
184North Mini Code82.7%estimated ± 5.0 pp, medium confidence
185GLM-4.582.6%estimated ± 4.8 pp, high confidence
186Gemma 4 26B A4B82.6%measured
187Quasar 438B82.6%estimated ± 7.0 pp, medium confidence
188Mistral Medium 382.6%estimated ± 5.0 pp, medium confidence
189Nemotron Ultra 253B82.5%estimated ± 4.4 pp, high confidence
190Qwen3-Omni-30B-A3B-Thinking82.4%estimated ± 7.0 pp, medium confidence
191GPT-4.182.3%estimated ± 4.0 pp, high confidence
192GLM-4.7-Flash82.3%estimated ± 4.4 pp, high confidence
193Claude 4 Opus82.2%estimated ± 4.4 pp, high confidence
194MiniMax M2.782.2%estimated ± 4.8 pp, high confidence
195LFM2.5-24B-A2B82.1%estimated ± 4.4 pp, high confidence
196Claude Opus 4.682.0%measured
197Claude 3 Haiku81.9%estimated ± 5.0 pp, medium confidence
198GPT-OSS 120B81.9%estimated ± 4.0 pp, high confidence
199A.X K281.9%estimated ± 5.0 pp, medium confidence
200Mistral Large 381.8%estimated ± 4.8 pp, high confidence
201Exaone 4.0 32B81.8%measured
202Qwen3 32B (Reasoning)81.8%estimated ± 1.1 pp, high confidence
203Grok 4 Fast (Reasoning)81.7%estimated ± 4.8 pp, high confidence
204Sarvam 105B81.7%measured
205Nemotron 3.5 Lightning 30B A3B NVFP481.6%measured
206Grok 4.1 Fast81.5%estimated ± 5.0 pp, medium confidence
207Nova Pro81.2%estimated ± 5.0 pp, medium confidence
208Claude Haiku 4.581.1%estimated ± 4.8 pp, high confidence
209Kimi K281.1%measured
210Llama 4 Maverick81.0%estimated ± 7.0 pp, medium confidence
211Magistral Medium81.0%estimated ± 4.4 pp, high confidence
212K-Exaone80.6%estimated ± 5.0 pp, medium confidence
213o1-preview80.5%estimated ± 5.4 pp, medium confidence
214GLM-4.5-Air80.5%estimated ± 5.0 pp, medium confidence
215QwQ-32B80.4%estimated ± 1.1 pp, high confidence
216Gemini 2.0 Flash80.2%estimated ± 3.6 pp, high confidence
217Solar Pro 280.2%estimated ± 5.0 pp, medium confidence
218Magistral Small80.1%estimated ± 4.4 pp, high confidence
219Claude 3.7 Sonnet80.0%estimated ± 4.4 pp, high confidence
220Sarvam 30B80.0%measured
221Grok 3 [Beta]79.9%measured
222Trinity-Large-Thinking79.5%estimated ± 5.4 pp, medium confidence
223LongCat-Flash-Lite-Sparse79.2%measured
224Qwen3 235B-A22B79.2%estimated ± 1.1 pp, high confidence
225Qwen3-Omni-30B-A3B-Instruct79.2%estimated ± 7.0 pp, medium confidence
226Claude Sonnet 4.679.2%measured
227Ternary Bonsai 2 27B79.0%estimated ± 1.1 pp, high confidence
228Gemini 1.5 Pro79.0%estimated ± 4.8 pp, high confidence
229Mistral Medium 3.5 128B79.0%estimated ± 4.8 pp, high confidence
230DeepSeek R1 Distill Llama 70B78.9%estimated ± 4.4 pp, high confidence
231Nemotron 3 Nano 30B78.3%measured
232LFM3-1.7B78.0%estimated ± 4.4 pp, high confidence
233Devstral Small78.0%estimated ± 5.0 pp, medium confidence
234Llama 4 Scout77.7%estimated ± 7.0 pp, medium confidence
235Granite 4.2 30B77.6%measured
236Nemotron 3 Nano Omni 30B A3B77.3%measured
237GPT-4o77.3%estimated ± 4.8 pp, high confidence
238Gemma 4 12B77.2%measured
239Mellum2.1-12B-A2.5B-Thinking77.1%estimated ± 1.1 pp, high confidence
240DeepSeek-R1-0528-Qwen3-8B76.8%estimated ± 4.4 pp, high confidence
241GPT-4.1 mini76.7%estimated ± 4.0 pp, high confidence
242Claude 3 Opus76.6%estimated ± 5.4 pp, medium confidence
243o4-mini76.5%estimated ± 3.6 pp, high confidence
244DeepSeek R1 Distill Qwen 32B76.2%estimated ± 5.4 pp, medium confidence
245Qwen2.5-Max76.1%measured
246LFM2.5-2.6B76.1%estimated ± 7.0 pp, medium confidence
247Trinity-Large-Preview76.1%estimated ± 4.0 pp, high confidence
248Phi-4-reasoning-plus76.0%measured
249Celeris-175.9%measured
250DeepSeek V375.9%measured
251Claude 3.5 Sonnet75.9%estimated ± 4.4 pp, high confidence
252MiniMax-Text-0175.7%measured
253DeepSeek R1 Distill Qwen 14B75.7%estimated ± 4.4 pp, high confidence
254Ling 2.6 Flash75.6%estimated ± 4.4 pp, high confidence
255Grok 275.5%measured
256o3-mini75.5%estimated ± 4.0 pp, high confidence
257GPT-4 Turbo75.2%estimated ± 5.4 pp, medium confidence
258Qwen2.5 Coder 32B Instruct75.1%estimated ± 5.4 pp, medium confidence
259DeepSeek-V2.575.0%estimated ± 5.4 pp, medium confidence
260Laguna XS.274.9%estimated ± 4.8 pp, high confidence
261Gemma 3 27B74.8%estimated ± 5.0 pp, medium confidence
262Laguna M.174.7%estimated ± 4.8 pp, high confidence
263Gemini 1.0 Pro74.7%estimated ± 5.4 pp, medium confidence
264Mellum2-12B-A2.5B-Thinking74.7%estimated ± 1.1 pp, high confidence
265Phi-4-reasoning74.3%measured
266ZAYA1-8B74.2%measured
267Gemini 2.0 Flash-Lite74.1%estimated ± 3.6 pp, high confidence
268Granite 4.2 8B74.0%measured
269Qwen3 32B73.9%estimated ± 1.1 pp, high confidence
270Particle-173.8%estimated ± 4.0 pp, high confidence
271Llama 3.1 405B73.1%estimated ± 7.0 pp, medium confidence
272GPT-OSS 20B72.3%estimated ± 4.0 pp, high confidence
273Grok 2 mini72.0%measured
274Ultravox v0.6 Llama 3.3 70B71.6%estimated ± 7.0 pp, medium confidence
275Phi-471.4%estimated ± 4.0 pp, high confidence
276MiniCPM5-2B70.8%measured
277Command A+69.9%estimated ± 5.0 pp, medium confidence
278LFM2.5-8B-A1B69.5%estimated ± 5.0 pp, medium confidence
279Gemma 4 E4B69.4%measured
280Llama 3.3 70B68.9%measured
281Qwen2.5-VL-32B68.8%measured
282ZAYA1-74B-Preview68.1%measured
283Granite 4.2 3B67.8%measured
284Gemini 2.5 Flash-Lite67.8%estimated ± 3.6 pp, high confidence
285Mistral Small 3.166.8%measured
286Llama 3.1 70B66.4%measured
287GPT-4o mini66.2%estimated ± 4.0 pp, high confidence
288Claude 3.5 Haiku65.0%measured
289Gemini 1.5 Flash64.0%estimated ± 3.6 pp, medium confidence
290Exaone 4.0 1.2B63.2%estimated ± 7.0 pp, medium confidence
291GPT-4.1 nano63.0%estimated ± 4.0 pp, high confidence
292Mellum2-12B-A2.5B-Instruct62.4%estimated ± 1.1 pp, high confidence
293o1-mini61.7%estimated ± 3.6 pp, medium confidence
294Gemma 4 E2B60.0%measured
295Mixtral 8x22B Instruct v0.159.1%estimated ± 4.0 pp, high confidence
296Gemma 2 27B56.6%estimated ± 4.0 pp, high confidence
297LLaDA2.2-mini55.5%estimated ± 5.4 pp, medium confidence
298Soofi S 30B-A3B51.4%measured
299MiniCPM5-1B48.9%measured
300Llama 3.1 8B48.3%measured
301Phi-3.5-mini47.4%measured
302Phi-4 Multimodal Instruct42.2%estimated ± 7.0 pp, medium confidence
303LFM2.5-VL-1.6B-Extract35.8%estimated ± 7.0 pp, medium confidence
304Granite-4.0-1B34.0%measured
305Granite-4.0-H-1B32.9%measured
306LFM2.5-230M20.3%measured
307LFM2.5-350M20.0%measured
308LFM2.5-VL-450M19.3%measured
309Granite-4.0-H-350M14.4%measured
310Granite-4.0-350M12.1%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General