Knowledge

MMLU-Pro Leaderboard

MMLU-Pro is a harder, cleaned-up successor to MMLU with ten answer choices and more reasoning-heavy questions across 14 subjects, measuring broad knowledge and reasoning together.

Source: tigerlab144 open models ranked+113 proprietary

All models ranked on MMLU-Pro

Proprietary / closed models are shown dimmed — you can't run them locally, but they show where the open field stands.

#ModelScore
1Gemini-3.1-Pro · proprietary
91.2%
2Gemini-3-Pro(11/25) · proprietary
90.1%
3GPT-o1 · proprietary
89.3%
4Claude-4.6-Opus(Thinking) · proprietary
89.1%
5Gemini-3-Flash(12/25) · proprietary
88.6%
6MiniMax M2.1 · 228.7B
88.0%
7Qwen3.5 397B A17B · 403.4B
87.8%
8Seed2.0-Lite · proprietary
87.7%
9GPT-5.4 · proprietary
87.5%
10Claude-4.5-Sonnet(Thinking) · proprietary
87.4%
11GPT-5.2 · proprietary
87.4%
12Claude-4-Opus-Thinking · proprietary
87.3%
13Claude-4.5-Opus(Thinking) · proprietary
87.3%
14Claude-4.6-Sonnet(Thinking) · proprietary
87.3%
15Hunyuan-T1 · proprietary
87.2%
16GPT-5(high) · proprietary
87.1%
17Kimi K2.5 · 1026.9B
87.1%
18Grok-4 · proprietary
87.0%
19Seed-Thinking-v1.5 · proprietary
87.0%
20Seed2.0-Pro · proprietary
87.0%
21Qwen3.5 122B A10B · 125.1B
86.7%
22Seed1.6-Base · proprietary
86.6%
23Seed1.6-Thinking · proprietary
86.6%
24GPT-5.1 · proprietary
86.4%
25Seed1.6-Ada-Thinking · proprietary
86.4%
26Gemini-3.1-Flash-Lite-Preview · proprietary
86.2%
27GPT-4.5 · proprietary
86.1%
28Qwen3.5 27B · 27.8B
86.1%
29Gemini-2.5-Pro · proprietary
86.0%
30GLM 5 · 753.9B
86.0%
31Qwen3-Max-Thinking · proprietary
85.7%
32Qwen3.5 35B A3B · 36.0B
85.3%
33DeepSeek V3.2 · 685.4B
85.0%
34GPT-o3-high · proprietary
85.0%
35DeepSeek V3.1 · 684.5B
84.8%
36GLM 4.5 · 358.3B
84.6%
37Gemini-2.5-Pro-Exp-03-25 · proprietary
84.5%
38Qwen3 235B A22B Thinking 2507 · 235.1B
84.5%
39Grok-4.1-Fast(Reasoning) · proprietary
84.2%
40Claude-3.7-Sonnet-Thinking · proprietary
84.0%
41DeepSeek R1 · 684.5B
84.0%
42Claude-4-Sonnet · proprietary
83.7%
43DeepSeek-V3.1-NonThinking · proprietary
83.7%
44Seed2.0-Mini · proprietary
83.6%
45Intern-S1 · proprietary
83.5%
46DeepSeek R1 0528 · 684.5B
83.4%
47GPT-4-mini (high) · proprietary
83.0%
48Grok-3-mini · proprietary
83.0%
49Qwen3 235B A22B Instruct 2507 · 235.1B
83.0%
50Llama4-Behemoth · proprietary
82.8%
51LongCat Flash Chat · 561.9B
82.7%
52Seed OSS 36B Instruct · 36.2B
82.7%
53Qwen3.5 9B · 9.7B
82.5%
54MiniMax M2 · 228.7B
82.0%
55GPT-4.1 · proprietary
81.8%
56GLM 4.5 Air · 110.5B
81.4%
57DeepSeek v3 0324 · 684.5B
81.3%
58MiniMax-M1 · proprietary
81.1%
59Kimi K2 Instruct · 1026.4B
81.0%
60Qwen3 30B A3B Thinking 2507 · 30.5B
80.9%
61GPT OSS 120B · 116.8B
80.8%
62Llama 4 Maverick 17B 128E Instruct · 401.6B
80.5%
63GPT-o1-mini · proprietary
80.3%
64Doubao-1.5-Pro · proprietary
80.1%
65MiniMax M2.5 · 228.7B
80.1%
66Grok3-Beta · proprietary
79.9%
67GPT-o3-mini · proprietary
79.4%
68Gemini-2.0-Pro · proprietary
79.1%
69Qwen3.5 4B · 4.7B
79.1%
70HunyuanTurboS · proprietary
79.0%
71Grok3-mini-Beta · proprietary
78.9%
72Qwen3-30B-A3B-Thinking · proprietary
78.5%
73ERNIE-4.5-300B-A47B · proprietary
78.4%
74NVIDIA Nemotron 3 Nano 30B A3B BF16 · 31.6B
78.3%
75Claude-3.5-Sonnet (2024-10-22) · proprietary
78.0%
76GPT-4o (2024-11-20) · proprietary
77.9%
77Gemini-2.0-Flash · proprietary
77.6%
78Gemini-2.0-Flash-exp · proprietary
76.2%
79Claude-3.5-Sonnet (2024-06-20) · proprietary
76.1%
80Qwen2.5-Max · proprietary
76.1%
81Phi 4 Reasoning Plus · 14.7B
76.0%
82DeepSeek v3 · 684.5B
75.9%
83MiniMax Text 01 · 456.1B
75.7%
84Grok-2 · proprietary
75.5%
85Grok-4.1-Fast(Non-Reasoning) · proprietary
75.2%
86GPT-4o (2024-08-06) · proprietary
74.7%
87Llama 4 Scout 17B 16E Instruct · 108.6B
74.3%
88Phi 4 Reasoning · 14.7B
74.3%
89GPT OSS 20B · 20.9B
73.6%
90Llama 3.1 405B Instruct · 405.9B
73.3%
91Athene-V2-Chat (0-shot) · proprietary
73.1%
92GPT-4o (2024-05-13) · proprietary
72.5%
93Grok-2-mini · proprietary
71.9%
94Gemini-2.0-Flash-Lite · proprietary
71.6%
95Qwen2.5 72B Instruct · 72.7B
71.6%
96ECHO_Ego_v2_14B · proprietary
71.2%
97QwQ 32B Preview · 32.8B
71.0%
98Phi 4 · 14.7B
70.4%
99Gemini-1.5-Pro-002 · proprietary
70.3%
100Athene v2 Chat · 72.7B
70.2%
101ERNIE-4.5-300B-A47B-Base · proprietary
69.5%
102Qwen2.5 32B Instruct · 32.8B
69.2%
103SkyThought-T1 · proprietary
69.2%
104QwQ 32B · 32.8B
69.1%
105Gemini-1.5-Pro · proprietary
69.0%
106Claude-3-Opus · proprietary
68.5%
107Qwen3 235B A22B · 235.1B
68.2%
108Mistral Large Instruct 2411 · 122.6B
67.9%
109Gemma 3 27B IT · 27.4B
67.5%
110Hunyuan-A13B · proprietary
67.3%
111Mistral-3.1-Small · proprietary
66.8%
112General-Reasoner-14B · proprietary
66.6%
113Mistral-Small-instruct · proprietary
66.3%
114Llama 3.3 70B Instruct · 70.6B
65.9%
115Mistral Large Instruct 2407 · 122.6B
65.9%
116DeepSeek V2.5 · 235.7B
65.8%
117NVIDIA Nemotron 3 Nano 30B A3B Base BF16 · 31.6B
65.1%
118Seed OSS 36B Base · 36.2B
65.1%
119Reka 3 · proprietary
65.0%
120Qwen2 72B Instruct · 72.7B
64.4%
121Gemini-1.5-Flash-002 · proprietary
64.1%
122magnum-72b-v1 · proprietary
63.9%
123GPT-4-Turbo · proprietary
63.7%
124Qwen2.5 14B · 14.8B
63.7%
125DeepSeek Coder v2 Instruct · 235.7B
63.6%
126Higgs Llama 3 70B · 70.6B
63.2%
127GPT-4o-mini · proprietary
63.1%
128azerogpt · proprietary
63.1%
129Llama 3.1 70B Instruct · 70.6B
62.8%
130Llama 3.1 Nemotron 70B Instruct HF · 70.6B
62.8%
131Yi-Lightning · proprietary
62.4%
132Claude-3-5-Haiku-20241022 · proprietary
62.1%
133RRD2.5-9B · proprietary
61.8%
134Qwen3 30B A3B Base · 30.5B
61.7%
135Llama 3.1 405B · 405.9B
61.6%
136Gemma 3 12B IT · 12.2B
60.6%
137Nemotron-H-56B-Base · proprietary
60.5%
138Reflection Llama 3.1 70B · 70.6B
60.4%
139Hunyuan-Large · proprietary
60.2%
140Gemini-1.5-Flash · proprietary
59.1%
141EXAONE 3.5 32B Instruct · 32.0B
58.9%
142General-Reasoner-7B · proprietary
58.9%
143MiMo 7B RL · 7.8B
58.6%
144Yi-large · proprietary
58.1%
145NewenAI/Phi4-sft · proprietary
57.7%
146Internlm3 8B Instruct · 8.8B
57.6%
147Claude-3-Sonnet · proprietary
56.8%
148ERNIE-4.5-21B-A3B-Base · proprietary
56.7%
149Gemma 2 27B IT · 27.2B
56.5%
150Mixtral 8x22B Instruct v0.1 · 140.6B
56.3%
151Meta Llama 3 70B Instruct · 70.6B
56.2%
152Phi 3 Medium 4k Instruct · 14.0B
55.7%
153Qwen2.5-Turbo · proprietary
55.6%
154Qwen2-72B-32k · proprietary
55.6%
155Qwen3.5 2B · 2.3B
55.3%
156Deepseek-V2-Chat · proprietary
54.8%
157Mistral-Small-base · proprietary
54.4%
158Phi 4 Mini Instruct · 3.8B
52.8%
159Meta Llama 3 70B · 70.6B
52.8%
160Qwen1.5 72B Chat · 72.3B
52.6%
161Llama 3.1 70B · 70.6B
52.5%
162Yi 1.5 34B Chat · 34.4B
52.3%
163Gemma 2 9B IT · 9.2B
52.1%
164Phi 3 Medium 128k Instruct · 14.0B
51.9%
165MAmmoTH2-8x7B-Plus · proprietary
50.4%
166Qwen1.5 110B · 111.2B
49.9%
167AI21 Jamba Large 1.5 · 398.6B
49.5%
168Mistral Small Instruct 2409 · 22.2B
48.4%
169Glm 4 9B Chat · 9.4B
48.0%
170GLM-4-9B · proprietary
47.9%
171Phi 3.5 Mini Instruct · 3.8B
47.9%
172Qwen2 7B Instruct · 7.6B
47.2%
173Cohere-Aya-Vision · proprietary
47.2%
174EXAONE 3.5 7.8B Instruct · 7.8B
46.2%
175Yi 1.5 9B Chat · 8.8B
46.0%
176Phi 3 Mini 4k Instruct · 3.8B
45.7%
177Aya Expanse 32B · 32.3B
45.4%
178Gemma 2 9B · 9.2B
45.1%
179Qwen2.5 7B · 7.6B
45.0%
180Mistral Nemo Instruct 2407 · 12.2B
44.8%
181Llama 3.1 8B Instruct · 8.0B
44.3%
182Nemotron H 8B Base 8K · 8.1B
44.0%
183Phi 3 Mini 128k Instruct · 3.8B
43.9%
184Qwen2.5 3B · 3.1B
43.7%
185Gemma 3 4B IT · 4.3B
43.6%
186MAmmoTH2-8B-Plus · proprietary
43.4%
187Mixtral 8x7B Instruct v0.1 · 46.7B
43.3%
188Yi 34B · 34.4B
43.0%
189Claude-3-Haiku-20240307 · proprietary
42.3%
190Mathstral 7B v0.1 · 7.2B
42.0%
191MiMo 7B Base · 7.8B
41.9%
192DeepSeek Coder v2 Lite Instruct · 15.7B
41.6%
193Granite 3.1 8B Instruct · 8.2B
41.0%
194Mixtral 8x7B v0.1 · 46.7B
41.0%
195Meta Llama 3 8B Instruct · 8.0B
41.0%
196MAmmoTH2-7B-Plus · proprietary
40.8%
197Qwen2 7B · 7.6B
40.7%
198Mistral Nemo Base 2407 · 12.2B
39.8%
199WizardLM 2 8x22B · 140.6B
39.2%
200EXAONE 3.5 2.4B Instruct · 2.4B
39.1%
201Yi 1.5 6B Chat · 6.1B
38.2%
202Qwen1.5 14B Chat · 14.2B
38.0%
203Ministral 8B Instruct 2410 · 8.0B
37.9%
204C4ai Command R V01 · 35.0B
37.9%
205Staring-7B · proprietary
37.9%
206Llama 2 70B HF · 69.0B
37.5%
207OpenChat-3.5-8B · proprietary
37.2%
208InternMath-20B-Plus · proprietary
37.1%
209LLaDA · proprietary
37.0%
210Llama3-Smaug-8B · proprietary
36.9%
211Llama 3.1 8B · 8.0B
36.6%
212Meta Llama 3 8B · 8.0B
35.4%
213DeepseekMath-7B-Instruct · proprietary
35.3%
214DeepSeek Coder v2 Lite Base · 15.7B
34.4%
215Aya Expanse 8B · 8.0B
33.7%
216Gemma 7B · 8.5B
33.7%
217InternMath-7B-Plus · proprietary
33.5%
218Granite-3.1-8B-Base · proprietary
33.1%
219Zephyr 7B Beta · 7.2B
33.0%
220Qwen2.5 1.5B · 1.5B
32.1%
221Granite 3.1 2B Instruct · 2.5B
32.0%
222Granite-3.0-8B-Base · proprietary
31.0%
223Mistral 7B v0.1 · 7.2B
30.9%
224Mistral 7B Instruct v0.2 · 7.2B
30.8%
225Mistral 7B v0.2 · 7.2B
30.4%
226Qwen3.5 0.8B · 873M
29.7%
227Qwen1.5 7B Chat · 7.7B
29.1%
228Yi 6B Chat · 6.1B
28.8%
229Neo-7B-Instruct · proprietary
28.7%
230Yi 6B · 6.1B
26.5%
231Neo-7B · proprietary
25.9%
232Mistral 7B Instruct v0.1 · 7.2B
25.8%
233Granite-3.1-3B-A800M-Instruct · proprietary
25.4%
234Llama 2 13B HF · 13.0B
25.3%
235Granite-3.1-2B-Base · proprietary
23.9%
236Llemma 7B · 7B
23.4%
237Qwen2 1.5B Instruct · 1.5B
22.6%
238Qwen2 1.5B · 1.5B
22.6%
239Llama 3.2 3B · 3.2B
22.2%
240Granite-3.0-2B-Base · proprietary
21.7%
241Granite-3.1-3B-A800M-Base · proprietary
20.4%
242Llama 2 7B HF · 6.7B
20.3%
243SmolLM2 1.7B · 1.7B
18.3%
244Qwen2 0.5B Instruct · 494M
15.9%
245Gemma 2B · 2.5B
15.8%
246Gemma 2 2B IT · 2.6B
15.6%
247Qwen2 0.5B · 494M
15.0%
248Qwen2.5 0.5B · 494M
14.9%
249Gemma 3 1B IT · 1000M
14.7%
250Granite-3.1-1B-A400M-Instruct · proprietary
13.3%
251Granite 3.1 1B A400m Base · 1.3B
12.3%
252Llama 3.2 1B · 1.2B
11.9%
253SmolLM 1.7B · 1.7B
11.9%
254SmolLM2 360M · 362M
11.4%
255SmolLM 135M · 135M
11.2%
256SmolLM 360M · 362M
10.9%
257SmolLM2 135M · 135M
10.8%

Score vs model size

Which models give the most quality for their size — the ones worth running locally.

1B10B100B1Tmodel size (log scale) →88.0%10.8%Qwen3.5 397B A17B · 403B · 87.8%Kimi K2.5 · 1T · 87.1%GLM 5 · 754B · 86.0%Qwen3.5 35B A3B · 36B · 85.3%DeepSeek V3.2 · 685B · 85.0%DeepSeek V3.1 · 685B · 84.8%GLM 4.5 · 358B · 84.6%Qwen3 235B A22B Thinking 2507 · 235B · 84.5%DeepSeek R1 · 684B · 84.0%DeepSeek R1 0528 · 685B · 83.4%Qwen3 235B A22B Instruct 2507 · 235B · 83.0%Seed OSS 36B Instruct · 36B · 82.7%LongCat Flash Chat · 562B · 82.7%MiniMax M2 · 229B · 82.0%GLM 4.5 Air · 110B · 81.4%DeepSeek v3 0324 · 685B · 81.3%Kimi K2 Instruct · 1T · 81.0%Qwen3 30B A3B Thinking 2507 · 31B · 80.9%GPT OSS 120B · 117B · 80.8%Llama 4 Maverick 17B 128E Instruct · 402B · 80.5%MiniMax M2.5 · 229B · 80.1%NVIDIA Nemotron 3 Nano 30B A3B BF16 · 32B · 78.3%Phi 4 Reasoning Plus · 15B · 76.0%DeepSeek v3 · 685B · 75.9%MiniMax Text 01 · 456B · 75.7%Llama 4 Scout 17B 16E Instruct · 109B · 74.3%Phi 4 Reasoning · 15B · 74.3%GPT OSS 20B · 21B · 73.6%Llama 3.1 405B Instruct · 406B · 73.3%Qwen2.5 72B Instruct · 73B · 71.6%QwQ 32B Preview · 33B · 71.0%Phi 4 · 15B · 70.4%Athene v2 Chat · 73B · 70.2%Qwen2.5 32B Instruct · 33B · 69.2%QwQ 32B · 33B · 69.1%Qwen3 235B A22B · 235B · 68.2%Mistral Large Instruct 2411 · 123B · 67.9%Gemma 3 27B IT · 27B · 67.5%Llama 3.3 70B Instruct · 71B · 65.9%Mistral Large Instruct 2407 · 123B · 65.9%DeepSeek V2.5 · 236B · 65.8%Seed OSS 36B Base · 36B · 65.1%NVIDIA Nemotron 3 Nano 30B A3B Base BF16 · 32B · 65.1%Qwen2 72B Instruct · 73B · 64.4%Qwen2.5 14B · 15B · 63.7%DeepSeek Coder v2 Instruct · 236B · 63.6%Higgs Llama 3 70B · 71B · 63.2%Llama 3.1 70B Instruct · 71B · 62.8%Llama 3.1 Nemotron 70B Instruct HF · 71B · 62.8%Qwen3 30B A3B Base · 31B · 61.7%Llama 3.1 405B · 406B · 61.6%Gemma 3 12B IT · 12B · 60.6%Reflection Llama 3.1 70B · 71B · 60.4%EXAONE 3.5 32B Instruct · 32B · 58.9%MiMo 7B RL · 8B · 58.6%Internlm3 8B Instruct · 9B · 57.6%Gemma 2 27B IT · 27B · 56.5%Mixtral 8x22B Instruct v0.1 · 141B · 56.3%Meta Llama 3 70B Instruct · 71B · 56.2%Phi 3 Medium 4k Instruct · 14B · 55.7%Phi 4 Mini Instruct · 4B · 52.8%Meta Llama 3 70B · 71B · 52.8%Qwen1.5 72B Chat · 72B · 52.6%Llama 3.1 70B · 71B · 52.5%Yi 1.5 34B Chat · 34B · 52.3%Gemma 2 9B IT · 9B · 52.1%Phi 3 Medium 128k Instruct · 14B · 51.9%Qwen1.5 110B · 111B · 49.9%AI21 Jamba Large 1.5 · 399B · 49.5%Mistral Small Instruct 2409 · 22B · 48.4%Glm 4 9B Chat · 9B · 48.0%Phi 3.5 Mini Instruct · 4B · 47.9%Qwen2 7B Instruct · 8B · 47.2%EXAONE 3.5 7.8B Instruct · 8B · 46.2%Yi 1.5 9B Chat · 9B · 46.0%Phi 3 Mini 4k Instruct · 4B · 45.7%Aya Expanse 32B · 32B · 45.4%Gemma 2 9B · 9B · 45.1%Qwen2.5 7B · 8B · 45.0%Mistral Nemo Instruct 2407 · 12B · 44.8%Llama 3.1 8B Instruct · 8B · 44.3%Nemotron H 8B Base 8K · 8B · 44.0%Phi 3 Mini 128k Instruct · 4B · 43.9%Qwen2.5 3B · 3B · 43.7%Gemma 3 4B IT · 4B · 43.6%Mixtral 8x7B Instruct v0.1 · 47B · 43.3%Yi 34B · 34B · 43.0%Mathstral 7B v0.1 · 7B · 42.0%MiMo 7B Base · 8B · 41.9%DeepSeek Coder v2 Lite Instruct · 16B · 41.6%Granite 3.1 8B Instruct · 8B · 41.0%Mixtral 8x7B v0.1 · 47B · 41.0%Meta Llama 3 8B Instruct · 8B · 41.0%Qwen2 7B · 8B · 40.7%Mistral Nemo Base 2407 · 12B · 39.8%WizardLM 2 8x22B · 141B · 39.2%EXAONE 3.5 2.4B Instruct · 2B · 39.1%Yi 1.5 6B Chat · 6B · 38.2%Qwen1.5 14B Chat · 14B · 38.0%Ministral 8B Instruct 2410 · 8B · 37.9%C4ai Command R V01 · 35B · 37.9%Llama 2 70B HF · 69B · 37.5%Llama 3.1 8B · 8B · 36.6%Meta Llama 3 8B · 8B · 35.4%DeepSeek Coder v2 Lite Base · 16B · 34.4%Aya Expanse 8B · 8B · 33.7%Gemma 7B · 9B · 33.7%Zephyr 7B Beta · 7B · 33.0%Granite 3.1 2B Instruct · 3B · 32.0%Mistral 7B v0.1 · 7B · 30.9%Mistral 7B Instruct v0.2 · 7B · 30.8%Mistral 7B v0.2 · 7B · 30.4%Qwen1.5 7B Chat · 8B · 29.1%Yi 6B Chat · 6B · 28.8%Yi 6B · 6B · 26.5%Mistral 7B Instruct v0.1 · 7B · 25.8%Llama 2 13B HF · 13B · 25.3%Llemma 7B · 7B · 23.4%Qwen2 1.5B Instruct · 2B · 22.6%Qwen2 1.5B · 2B · 22.6%Llama 3.2 3B · 3B · 22.2%Llama 2 7B HF · 7B · 20.3%SmolLM2 1.7B · 2B · 18.3%Gemma 2B · 3B · 15.8%Gemma 2 2B IT · 3B · 15.6%Qwen2 0.5B · 494M · 15.0%Qwen2.5 0.5B · 494M · 14.9%Gemma 3 1B IT · 1000M · 14.7%Granite 3.1 1B A400m Base · 1B · 12.3%Llama 3.2 1B · 1B · 11.9%SmolLM 1.7B · 2B · 11.9%SmolLM 360M · 362M · 10.9%SmolLM2 135M · 135M · 10.8%SmolLM 135M · 135M · 11.2%SmolLM 135MSmolLM2 360M · 362M · 11.4%SmolLM2 360MQwen2 0.5B Instruct · 494M · 15.9%Qwen2 0.5B InstructQwen3.5 0.8B · 873M · 29.7%Qwen3.5 0.8BQwen2.5 1.5B · 2B · 32.1%Qwen2.5 1.5BQwen3.5 2B · 2B · 55.3%Qwen3.5 2BQwen3.5 4B · 5B · 79.1%Qwen3.5 4BQwen3.5 9B · 10B · 82.5%Qwen3.5 9BQwen3.5 27B · 28B · 86.1%Qwen3.5 27BQwen3.5 122B A10B · 125B · 86.7%Qwen3.5 122B A10BMiniMax M2.1 · 229B · 88.0%MiniMax M2.1
Each dot is a model. Up = higher score, left = smaller (easier to run locally). The dashed line marks the efficiency frontier — the best score you can get at each size or smaller.
  • SmolLM 135M, 135M, score 11.2% — on the efficiency frontier (best score at its size or smaller).
  • SmolLM2 360M, 362M, score 11.4% — on the efficiency frontier (best score at its size or smaller).
  • Qwen2 0.5B Instruct, 494M, score 15.9% — on the efficiency frontier (best score at its size or smaller).
  • Qwen3.5 0.8B, 873M, score 29.7% — on the efficiency frontier (best score at its size or smaller).
  • Qwen2.5 1.5B, 2B, score 32.1% — on the efficiency frontier (best score at its size or smaller).
  • Qwen3.5 2B, 2B, score 55.3% — on the efficiency frontier (best score at its size or smaller).
  • Qwen3.5 4B, 5B, score 79.1% — on the efficiency frontier (best score at its size or smaller).
  • Qwen3.5 9B, 10B, score 82.5% — on the efficiency frontier (best score at its size or smaller).
  • Qwen3.5 27B, 28B, score 86.1% — on the efficiency frontier (best score at its size or smaller).
  • Qwen3.5 122B A10B, 125B, score 86.7% — on the efficiency frontier (best score at its size or smaller).
  • MiniMax M2.1, 229B, score 88.0% — on the efficiency frontier (best score at its size or smaller).

MMLU-Pro: frequently asked questions

What is the best open LLM on MMLU-Pro?
MiniMax M2.1 is the top open model on MMLU-Pro, scoring 88.0%. Among all models tested — including proprietary ones — it ranks #6. The top model overall is Gemini-3.1-Pro (Google) at 91.2%.
What's the best MMLU-Pro model you can run on a 24 GB GPU?
Qwen3.5 27B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (about 15 GB), scoring 86.1% on MMLU-Pro.
What's the best MMLU-Pro model you can run on a 12 GB GPU?
Qwen3.5 9B is the highest-scoring open model that fits in 12 GB at 4-bit quantization (about 5 GB), scoring 82.5% on MMLU-Pro.
Can open models match proprietary models on MMLU-Pro?
Not quite on MMLU-Pro: the strongest proprietary model (Gemini-3.1-Pro) scores 91.2%, ahead of the best open model (MiniMax M2.1) at 88.0% — but you can run the open one yourself.

Scores aggregated from tigerlab. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.