benchgap
Knowledge & reasoning

MMMLU leaderboard

As of 2026-11-21, the highest measured score on MMMLU is 90.9% by Interfaze Beta. 275 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.295.5%estimated ± 2.6 pp, low confidence
2Qwen4 Max94.6%estimated ± 3.0 pp, low confidence
3Claude Opus 5.593.4%estimated ± 2.6 pp, low confidence
4Gemini 4 Argon92.7%estimated ± 2.6 pp, low confidence
5Claude Sonnet 5.592.4%estimated ± 2.6 pp, low confidence
6Muse Ember92.4%estimated ± 3.0 pp, low confidence
7Muse Wonder Full92.1%estimated ± 2.6 pp, low confidence
8GPT-6.1 Sol92.0%estimated ± 2.6 pp, low confidence
9GPT-6.2 Astra91.8%estimated ± 1.1 pp, low confidence
10Muse Wonder91.6%estimated ± 2.6 pp, low confidence
11GPT-6.2 Sol91.6%estimated ± 1.1 pp, low confidence
12GPT-6 Astra91.6%estimated ± 1.1 pp, low confidence
13Claude Mythos 591.5%estimated ± 1.7 pp, medium confidence
14Sakana Fugu91.4%estimated ± 1.1 pp, low confidence
15Sakana Fugu-Ultra91.4%estimated ± 1.1 pp, low confidence
16MiMo-V2.6-Pro91.4%estimated ± 0.7 pp, low confidence
17Muse Spark 1.391.3%estimated ± 0.7 pp, low confidence
18Kimi K3.1 Pro91.2%estimated ± 0.7 pp, low confidence
19GPT-6 Sol91.2%estimated ± 2.6 pp, low confidence
20GPT-5.6 Sol91.2%estimated ± 1.1 pp, low confidence
21GPT-5.491.1%estimated ± 0.7 pp, low confidence
22Kimi K391.1%estimated ± 0.7 pp, low confidence
23MiniMax M3.1 Flash91.0%estimated ± 0.7 pp, low confidence
24GPT-5.590.9%estimated ± 1.1 pp, low confidence
25GPT-6.1 Terra90.9%estimated ± 2.6 pp, low confidence
26Interfaze Beta90.9%measured
27Claude Opus 4.890.9%estimated ± 0.7 pp, low confidence
28GLM-5.290.9%estimated ± 0.7 pp, low confidence
29Step 5 Preview90.9%estimated ± 0.7 pp, low confidence
30GPT-5.6 Luna90.9%estimated ± 0.7 pp, low confidence
31Sage-A1.590.8%estimated ± 0.7 pp, low confidence
32Inkling 2 Small90.8%estimated ± 1.1 pp, low confidence
33GPT-5.6 Terra90.8%estimated ± 1.1 pp, low confidence
34GPT-6 Luna90.7%estimated ± 0.7 pp, low confidence
35Ornith-1.5-397B90.7%estimated ± 1.1 pp, low confidence
36GLM-5.390.7%estimated ± 0.7 pp, low confidence
37Claude Haiku 5.590.7%estimated ± 0.7 pp, low confidence
38Qwen3.8 Max90.7%estimated ± 1.1 pp, low confidence
39Gemini 3.8 Flash90.6%estimated ± 0.7 pp, low confidence
40Pareto 26.10 Preview90.6%estimated ± 1.1 pp, medium confidence
41DeepSeek V4 Pro90.6%estimated ± 0.7 pp, low confidence
42Ling 3.1 Flash90.6%estimated ± 0.7 pp, low confidence
43Gemini 3.1 Pro90.6%estimated ± 0.7 pp, low confidence
44Grok 4.790.6%estimated ± 0.7 pp, low confidence
45Muse Spark 1.290.6%estimated ± 0.7 pp, low confidence
46Qwen3.8 Max Preview90.6%estimated ± 0.7 pp, low confidence
47Beam 1.0 501B-A23B FP890.5%estimated ± 0.7 pp, low confidence
48Grok 4.690.5%estimated ± 0.7 pp, low confidence
49Claude Sonnet 590.5%estimated ± 0.7 pp, low confidence
50GPT-5.3 Codex90.5%estimated ± 0.7 pp, low confidence
51Hy4 preview90.5%estimated ± 0.7 pp, low confidence
52DeepSeek V4 Flash 073190.4%estimated ± 0.7 pp, low confidence
53Beam 1.0 501B-A23B90.4%estimated ± 0.7 pp, low confidence
54Qwen3.7 Max90.3%measured
55Grok 4.590.3%estimated ± 0.7 pp, low confidence
56Beam 1.0 501B-A23B NVFP490.3%estimated ± 0.7 pp, low confidence
57Muse Spark 1.190.3%estimated ± 0.7 pp, low confidence
58Muse Spark Open90.2%estimated ± 0.7 pp, low confidence
59Qwen3.8-Omni-Flash90.2%estimated ± 1.1 pp, medium confidence
60DeepSeek V4.1 Flash 101790.2%estimated ± 0.7 pp, low confidence
61Gemini 3.7 Flash90.2%estimated ± 0.7 pp, low confidence
62Gemini 3.5 Flash90.0%estimated ± 0.7 pp, medium confidence
63Particle-190.0%estimated ± 1.1 pp, medium confidence
64Claude Opus 4.6 (Adaptive)89.9%estimated ± 0.7 pp, medium confidence
65Claude Fable 5.189.9%estimated ± 1.7 pp, medium confidence
66Claude Opus 4.789.8%estimated ± 0.7 pp, medium confidence
67MiMo-V2.6-Flash89.8%estimated ± 0.7 pp, medium confidence
68Claude Opus 589.8%estimated ± 1.7 pp, medium confidence
69GPT-5.289.8%estimated ± 0.7 pp, medium confidence
70Muse Spark89.7%estimated ± 0.7 pp, medium confidence
71Qwen3.8-Flash-Next89.7%estimated ± 0.7 pp, medium confidence
72Claude Fable 589.7%estimated ± 1.7 pp, medium confidence
73Nemotron 3.5 Super VL 120B A12B89.7%estimated ± 0.7 pp, medium confidence
74Gemini 3.6 Flash89.6%estimated ± 0.7 pp, medium confidence
75Nemotron 3.5 Super VL 120B A12B NVFP489.6%estimated ± 0.7 pp, medium confidence
76Mercury 389.5%estimated ± 1.1 pp, medium confidence
77Ornith-1.5-35B-A3B89.5%estimated ± 1.1 pp, medium confidence
78GPT-5.4 mini89.5%estimated ± 0.7 pp, medium confidence
79Kimi K2.7 Code89.5%estimated ± 0.7 pp, medium confidence
80Mistral Large 489.5%estimated ± 0.7 pp, medium confidence
81Quasar 438B89.4%estimated ± 0.7 pp, medium confidence
82Nemotron 3.5 Bolt 74B-A9B BF1689.4%estimated ± 1.1 pp, medium confidence
83GPT-5.4 nano89.4%estimated ± 0.7 pp, medium confidence
84Gemini 3 Pro89.3%estimated ± 0.7 pp, medium confidence
85A.X K289.3%estimated ± 0.7 pp, medium confidence
86GPT-5.2-Codex89.3%estimated ± 0.7 pp, medium confidence
87Grok 4.2089.3%estimated ± 1.1 pp, medium confidence
88Inkling-Small89.2%estimated ± 0.7 pp, medium confidence
89OrcaCyber-Nano-1.089.2%estimated ± 1.1 pp, medium confidence
90Grok 4.389.1%estimated ± 0.7 pp, medium confidence
91Kimi K2.689.1%estimated ± 0.7 pp, medium confidence
92Qwen3.7 Plus89.0%measured
93Claude 4 Opus Thinking88.8%measured
94Solar Pro 488.7%estimated ± 0.7 pp, medium confidence
95GPT-588.6%estimated ± 0.7 pp, medium confidence
96GPT-5.1-Codex88.6%estimated ± 0.7 pp, medium confidence
97GPT-5.1-Codex-Max88.6%estimated ± 0.7 pp, medium confidence
98Inkling88.5%estimated ± 0.7 pp, medium confidence
99Qwen3.8-27B88.5%estimated ± 0.7 pp, medium confidence
100Inkling 2 Mini88.5%estimated ± 1.1 pp, medium confidence
101Ornith-1.5-9B88.4%estimated ± 1.1 pp, medium confidence
102GPT-5.188.4%estimated ± 0.7 pp, medium confidence
103Hy388.4%estimated ± 0.7 pp, medium confidence
104Hy3 Preview88.4%estimated ± 0.7 pp, medium confidence
105Solar Open 288.3%estimated ± 1.1 pp, medium confidence
106Apodex 1.188.3%estimated ± 0.7 pp, medium confidence
107Apodex 1.1 Mini88.3%estimated ± 0.7 pp, medium confidence
108Claude Opus 4.5 Thinking88.3%estimated ± 0.7 pp, medium confidence
109GLM-5.188.3%estimated ± 0.7 pp, medium confidence
110Nemotron 3.5 Bolt 74B-A9B88.2%estimated ± 0.7 pp, medium confidence
111Claude Sonnet 4.588.1%estimated ± 1.7 pp, high confidence
112o3-pro88.1%estimated ± 1.7 pp, high confidence
113Ternary Bonsai 2 27B88.1%estimated ± 1.1 pp, medium confidence
114MiMo-V2.5-Pro88.1%estimated ± 0.7 pp, medium confidence
115MiniMax M387.9%estimated ± 0.7 pp, medium confidence
116Qwen 3.6 Max (preview)87.9%estimated ± 0.7 pp, medium confidence
117Ripple-65B Preview87.8%estimated ± 1.1 pp, medium confidence
118Kimi K2.587.7%estimated ± 0.7 pp, medium confidence
119Nemotron 3 Ultra87.7%estimated ± 0.7 pp, medium confidence
120Grok 4.1 Fast (Reasoning)87.6%estimated ± 0.7 pp, medium confidence
121Grok 4 Fast (Reasoning)87.6%estimated ± 0.7 pp, medium confidence
122Qwen3.6 Plus87.6%estimated ± 0.7 pp, medium confidence
123Claude Opus 4.687.6%estimated ± 0.7 pp, medium confidence
124Gemini 2.5 Pro87.5%estimated ± 0.7 pp, medium confidence
125Muse Glimmer 30B87.5%estimated ± 0.7 pp, medium confidence
126Claude 4 Opus87.4%measured
127Step 3.7 Flash87.3%estimated ± 0.7 pp, medium confidence
128MAI-Thinking-187.3%estimated ± 1.1 pp, medium confidence
129Ling 3.0 Flash FP887.2%estimated ± 1.1 pp, medium confidence
130GLM-587.2%estimated ± 0.7 pp, medium confidence
131Grok 487.2%estimated ± 0.7 pp, medium confidence
132Ling 3.0 Flash VL87.2%estimated ± 0.7 pp, medium confidence
133Grok 3 Mini87.1%estimated ± 1.7 pp, high confidence
134GLM-4.787.0%estimated ± 0.7 pp, medium confidence
135Ling 3.0 Flash87.0%estimated ± 0.7 pp, medium confidence
136Qwen3.5 397B87.0%estimated ± 0.7 pp, medium confidence
137Claude 4.1 Opus86.8%estimated ± 3.0 pp, medium confidence
138DeepSeek-R186.8%estimated ± 0.7 pp, medium confidence
139Gemini 2.5 Flash86.8%estimated ± 0.7 pp, medium confidence
140Gemini 3 Flash86.8%estimated ± 0.7 pp, medium confidence
141Gemma 4 31B86.8%estimated ± 0.7 pp, medium confidence
142GPT-OSS 20B86.8%estimated ± 0.7 pp, medium confidence
143Nemotron 3 Super 100B86.7%estimated ± 1.7 pp, high confidence
144K-EXAONE 2.086.6%measured
145o1-pro86.6%estimated ± 1.7 pp, high confidence
146Claude 4 Sonnet86.5%estimated ± 0.7 pp, medium confidence
147GPT-OSS 120B86.5%estimated ± 0.7 pp, medium confidence
148K-Exaone86.5%estimated ± 0.7 pp, medium confidence
149MiMo-V2-Omni86.5%estimated ± 0.7 pp, medium confidence
150o386.5%estimated ± 0.7 pp, medium confidence
151Qwen3.6-27B86.5%estimated ± 0.7 pp, medium confidence
152o1-preview86.4%estimated ± 1.7 pp, high confidence
153Claude Sonnet 4.686.3%estimated ± 0.7 pp, medium confidence
154DeepSeek V3.286.3%estimated ± 0.7 pp, medium confidence
155Qwen3.5-27B86.3%estimated ± 0.7 pp, medium confidence
156Qwen3.5-35B-A3B86.3%estimated ± 0.7 pp, medium confidence
157Trinity-Large-Preview86.3%estimated ± 0.7 pp, medium confidence
158Trinity-Large-Thinking86.3%estimated ± 0.7 pp, medium confidence
159DeepSeek-R1-052886.1%estimated ± 1.4 pp, medium confidence
160Claude 3.7 Sonnet Thinking86.1%measured
161Qwen3 235B 250786.0%estimated ± 1.7 pp, high confidence
162Gemma 4 E4B85.9%estimated ± 0.7 pp, medium confidence
163GLM-5V-Turbo85.9%estimated ± 0.7 pp, medium confidence
164MiniMax M2.785.9%estimated ± 0.7 pp, medium confidence
165o4-mini85.9%estimated ± 0.7 pp, medium confidence
166Qwen3.5-122B-A10B85.9%estimated ± 0.7 pp, medium confidence
167o3-mini85.9%estimated ± 1.7 pp, high confidence
168Claude Opus 4.585.4%estimated ± 0.7 pp, medium confidence
169Command A+85.4%estimated ± 0.7 pp, medium confidence
170GLM-5-Turbo85.4%estimated ± 0.7 pp, medium confidence
171Granite 4.2 30B85.4%estimated ± 0.7 pp, medium confidence
172Granite 4.2 8B85.4%estimated ± 0.7 pp, medium confidence
173Magistral Medium85.4%estimated ± 0.7 pp, medium confidence
174MiMo-V2-Pro85.4%estimated ± 0.7 pp, medium confidence
175MiniCPM5-2B85.4%estimated ± 0.7 pp, medium confidence
176Mistral Small 485.4%estimated ± 0.7 pp, medium confidence
177North Mini Code85.4%estimated ± 0.7 pp, medium confidence
178o185.4%estimated ± 0.7 pp, medium confidence
179Qwen3.6-35B-A3B85.4%estimated ± 0.7 pp, medium confidence
180Sarvam 30B85.4%estimated ± 0.7 pp, medium confidence
181Grok 3 [Beta]85.2%estimated ± 1.7 pp, high confidence
182GLM-4.7-Flash85.1%estimated ± 1.7 pp, high confidence
183Qwen3 235B-A22B (Reasoning)84.3%measured
184GPT-4 Turbo84.0%estimated ± 2.6 pp, low confidence
185Gemma 4 12B83.4%measured
186Celeris-183.3%estimated ± 0.7 pp, medium confidence
187Claude 3.5 Haiku83.3%estimated ± 0.7 pp, medium confidence
188Claude 3 Haiku83.3%estimated ± 0.7 pp, medium confidence
189DeepSeek V383.3%estimated ± 0.7 pp, medium confidence
190DeepSeek V3 032483.3%estimated ± 0.7 pp, medium confidence
191DeepSeek V3.183.3%estimated ± 0.7 pp, medium confidence
192Devstral Small83.3%estimated ± 0.7 pp, medium confidence
193Exaone 4.0 1.2B83.3%estimated ± 0.7 pp, medium confidence
194Exaone 4.0 32B83.3%estimated ± 0.7 pp, medium confidence
195Gemini 2.0 Flash83.3%estimated ± 0.7 pp, medium confidence
196Gemini 2.5 Flash-Lite83.3%estimated ± 0.7 pp, medium confidence
197Gemini 3.5 Flash-Lite83.3%estimated ± 0.7 pp, medium confidence
198Gemma 3 27B83.3%estimated ± 0.7 pp, medium confidence
199Gemma 4 26B A4B83.3%estimated ± 0.7 pp, medium confidence
200Gemma 4 E2B83.3%estimated ± 0.7 pp, medium confidence
201GLM-4.5-Air83.3%estimated ± 0.7 pp, medium confidence
202GLM-4.683.3%estimated ± 0.7 pp, medium confidence
203GPT-4.183.3%estimated ± 0.7 pp, medium confidence
204GPT-4.1 mini83.3%estimated ± 0.7 pp, medium confidence
205GPT-4.1 nano83.3%estimated ± 0.7 pp, medium confidence
206GPT-4o83.3%estimated ± 0.7 pp, medium confidence
207Granite-4.0-350M83.3%estimated ± 0.7 pp, medium confidence
208Granite-4.0-H-1B83.3%estimated ± 0.7 pp, medium confidence
209Granite-4.0-H-350M83.3%estimated ± 0.7 pp, medium confidence
210Granite 4.2 3B83.3%estimated ± 0.7 pp, medium confidence
211Grok 4.1 Fast83.3%estimated ± 0.7 pp, medium confidence
212Grok Code Fast 183.3%estimated ± 0.7 pp, medium confidence
213Kimi K283.3%estimated ± 0.7 pp, medium confidence
214LFM2.5-2.6B83.3%estimated ± 0.7 pp, medium confidence
215LFM2.5-8B-A1B83.3%estimated ± 0.7 pp, medium confidence
216LFM2.5-VL-1.6B-Extract83.3%estimated ± 0.7 pp, medium confidence
217Ling 2.6 Flash83.3%estimated ± 0.7 pp, medium confidence
218Ling 3.0 Tiny83.3%estimated ± 0.7 pp, medium confidence
219Llama 3.1 405B83.3%estimated ± 0.7 pp, medium confidence
220Llama 3.1 70B83.3%estimated ± 0.7 pp, medium confidence
221Llama 3.3 70B83.3%estimated ± 0.7 pp, medium confidence
222Llama 4 Maverick83.3%estimated ± 0.7 pp, medium confidence
223Llama 4 Scout83.3%estimated ± 0.7 pp, medium confidence
224Magistral Small83.3%estimated ± 0.7 pp, medium confidence
225Mercury 2.583.3%estimated ± 0.7 pp, medium confidence
226MiMo-V2-Flash83.3%estimated ± 0.7 pp, medium confidence
227Mistral Large 283.3%estimated ± 0.7 pp, medium confidence
228Mistral Large 383.3%estimated ± 0.7 pp, medium confidence
229Mistral Medium 383.3%estimated ± 0.7 pp, medium confidence
230Mistral Medium 3.5 128B83.3%estimated ± 0.7 pp, medium confidence
231Mistral Small 3.183.3%estimated ± 0.7 pp, medium confidence
232Nemotron 3.5 Lightning 30B A3B NVFP483.3%estimated ± 0.7 pp, medium confidence
233Nemotron 3 Nano 30B83.3%estimated ± 0.7 pp, medium confidence
234Nemotron 3 Nano Omni 30B A3B83.3%estimated ± 0.7 pp, medium confidence
235Nemotron Ultra 253B83.3%estimated ± 0.7 pp, medium confidence
236Nova Pro83.3%estimated ± 0.7 pp, medium confidence
237Phi-483.3%estimated ± 0.7 pp, medium confidence
238Phi-4-reasoning83.3%estimated ± 0.7 pp, medium confidence
239Qwen3.5-4B83.3%estimated ± 0.7 pp, medium confidence
240Qwen3 Max83.3%estimated ± 0.7 pp, medium confidence
241Qwen3-Omni-30B-A3B-Instruct83.3%estimated ± 0.7 pp, medium confidence
242Qwen3-Omni-30B-A3B-Thinking83.3%estimated ± 0.7 pp, medium confidence
243Sarvam 105B83.3%estimated ± 0.7 pp, medium confidence
244Solar Pro 283.3%estimated ± 0.7 pp, medium confidence
245Solar Pro 383.3%estimated ± 0.7 pp, medium confidence
246Ultravox v0.6 Llama 3.3 70B83.3%estimated ± 0.7 pp, medium confidence
247Claude 3.7 Sonnet83.2%measured
248GPT-4.582.8%estimated ± 1.7 pp, high confidence
249Phi-4-reasoning-plus82.5%estimated ± 1.7 pp, high confidence
250Spark-X2.6-4B82.3%estimated ± 1.1 pp, low confidence
251DeepSeek R1 Distill Qwen 32B82.1%estimated ± 1.7 pp, high confidence
252Gemini 1.5 Pro81.2%estimated ± 1.7 pp, medium confidence
253QwQ-32B80.9%measured
254DeepSeek R1 Distill Llama 70B80.9%estimated ± 1.7 pp, high confidence
255Qwen3 32B (Reasoning)80.6%measured
256LFM3-1.7B80.0%estimated ± 1.7 pp, high confidence
257Qwen3 235B-A22B79.8%measured
258LFM2.5-24B-A2B79.8%estimated ± 1.1 pp, low confidence
259DeepSeek-R1-0528-Qwen3-8B78.9%estimated ± 1.7 pp, high confidence
260Qwen2.5-Max78.3%estimated ± 1.7 pp, high confidence
261o1-mini78.3%estimated ± 1.7 pp, high confidence
262Mellum2.1-12B-A2.5B-Thinking78.0%estimated ± 1.4 pp, medium confidence
263Claude 3.5 Sonnet78.0%estimated ± 1.7 pp, high confidence
264DeepSeek R1 Distill Qwen 14B77.8%estimated ± 1.7 pp, high confidence
265Claude 3 Opus77.1%estimated ± 1.7 pp, medium confidence
266Mellum2-12B-A2.5B-Thinking77.1%estimated ± 1.4 pp, medium confidence
267ZAYA1-74B-Preview76.8%estimated ± 1.7 pp, high confidence
268Qwen3 32B76.5%measured
269Grok 276.1%estimated ± 1.7 pp, high confidence
270ZAYA1-8B75.7%estimated ± 1.1 pp, low confidence
271Mellum2-12B-A2.5B-Instruct75.4%estimated ± 1.4 pp, low confidence
272MiniMax-Text-0175.2%estimated ± 1.7 pp, high confidence
273GPT-4o mini73.8%estimated ± 1.7 pp, medium confidence
274DeepSeek-V2.573.6%estimated ± 1.7 pp, medium confidence
275LongCat-Flash-Lite-Sparse73.6%estimated ± 1.1 pp, low confidence
276Qwen2.5 Coder 32B Instruct73.3%estimated ± 1.7 pp, medium confidence
277Grok 2 mini73.2%estimated ± 1.7 pp, high confidence
278Gemini 2.0 Flash-Lite72.8%estimated ± 1.7 pp, high confidence
279Gemini 1.5 Flash72.5%estimated ± 1.7 pp, high confidence
280Qwen2.5-VL-32B69.8%estimated ± 1.7 pp, high confidence
281Soofi S 30B-A3B67.9%estimated ± 1.7 pp, high confidence
282Phi-4 Multimodal Instruct66.2%estimated ± 1.7 pp, medium confidence
283Gemini 1.0 Pro62.7%estimated ± 1.7 pp, medium confidence
284MiniCPM5-1B61.0%estimated ± 3.4 pp, medium confidence
285LFM2.5-350M56.7%estimated ± 1.7 pp, high confidence
286Llama 3.1 8B56.4%estimated ± 1.7 pp, high confidence
287Granite-4.0-1B55.9%estimated ± 1.7 pp, medium confidence
288Phi-3.5-mini55.4%measured
289LFM2.5-VL-450M51.1%estimated ± 1.7 pp, medium confidence
290LFM2.5-230M50.8%estimated ± 1.7 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General