benchgap
Knowledge & reasoning

MMLU-Pro (Arcee) leaderboard

As of 2026-11-21, the highest measured score on MMLU-Pro (Arcee) is 89.1% by Claude Opus 4.6. 187 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-5.195.5%estimated ± 4.0 pp, low confidence
2Hy395.5%estimated ± 4.0 pp, low confidence
3Apodex 1.194.4%estimated ± 4.0 pp, low confidence
4Apodex 1.1 Mini94.4%estimated ± 4.0 pp, low confidence
5Claude Opus 4.5 Thinking94.4%estimated ± 4.0 pp, low confidence
6MiMo-V2.5-Pro92.1%estimated ± 4.0 pp, low confidence
7MiniMax M391.0%estimated ± 4.0 pp, low confidence
8Qwen 3.6 Max (preview)91.0%estimated ± 4.0 pp, low confidence
9GPT-6.2 Astra89.1%estimated ± 3.4 pp, low confidence
10Claude Opus 4.689.1%measured
11GPT-6.2 Sol88.8%estimated ± 3.4 pp, low confidence
12GPT-6 Astra88.8%estimated ± 3.4 pp, low confidence
13Sakana Fugu88.7%estimated ± 3.4 pp, low confidence
14Sakana Fugu-Ultra88.7%estimated ± 3.4 pp, low confidence
15GPT-5.6 Sol88.4%estimated ± 3.4 pp, low confidence
16Gemini 3.1 Pro88.3%estimated ± 3.4 pp, low confidence
17Claude Opus 4.788.2%estimated ± 3.4 pp, low confidence
18Kimi K3.1 Pro88.2%estimated ± 3.4 pp, low confidence
19Claude Opus 4.888.1%estimated ± 3.4 pp, low confidence
20GPT-5.588.1%estimated ± 3.4 pp, low confidence
21Kimi K388.0%estimated ± 3.4 pp, low confidence
22Step 5 Preview88.0%estimated ± 3.4 pp, low confidence
23Grok 4.1 Fast (Reasoning)88.0%estimated ± 4.0 pp, medium confidence
24Grok 4 Fast (Reasoning)88.0%estimated ± 4.0 pp, medium confidence
25Qwen3.6 Plus88.0%estimated ± 4.0 pp, medium confidence
26Inkling 2 Small87.9%estimated ± 3.4 pp, low confidence
27GPT-5.6 Terra87.8%estimated ± 3.4 pp, low confidence
28GPT-5.487.8%estimated ± 3.4 pp, low confidence
29Ornith-1.5-397B87.8%estimated ± 3.4 pp, low confidence
30Gemini 3.5 Flash87.8%estimated ± 3.4 pp, low confidence
31Qwen3.8 Max87.7%estimated ± 3.4 pp, low confidence
32Pareto 26.10 Preview87.7%estimated ± 3.4 pp, low confidence
33Qwen3.7 Max87.7%estimated ± 3.4 pp, low confidence
34GPT-5.6 Luna87.6%estimated ± 3.4 pp, low confidence
35Hy4 preview87.6%estimated ± 3.4 pp, low confidence
36Qwen3.8-Flash-Next87.4%estimated ± 3.4 pp, low confidence
37GLM-5.287.3%estimated ± 3.4 pp, low confidence
38Qwen3.8-Omni-Flash87.2%estimated ± 3.4 pp, low confidence
39DeepSeek V4.1 Flash 101787.2%estimated ± 3.4 pp, low confidence
40Kimi K2.587.1%measured
41Beam 1.0 501B-A23B87.1%estimated ± 3.4 pp, low confidence
42Kimi K2.687.0%estimated ± 3.4 pp, low confidence
43Particle-187.0%estimated ± 3.4 pp, low confidence
44Qwen3.7 Plus87.0%estimated ± 3.4 pp, low confidence
45Beam 1.0 501B-A23B FP887.0%estimated ± 3.4 pp, low confidence
46DeepSeek V4 Pro86.9%estimated ± 3.4 pp, low confidence
47Gemini 2.5 Pro86.9%estimated ± 4.0 pp, medium confidence
48Muse Glimmer 30B86.9%estimated ± 4.0 pp, medium confidence
49Interfaze Beta86.8%estimated ± 3.4 pp, low confidence
50Nemotron 3.5 Super VL 120B A12B NVFP486.8%estimated ± 3.4 pp, low confidence
51Nemotron 3.5 Super VL 120B A12B86.7%estimated ± 3.4 pp, low confidence
52Inkling-Small86.7%estimated ± 3.4 pp, low confidence
53Muse Spark86.7%estimated ± 3.4 pp, low confidence
54Beam 1.0 501B-A23B NVFP486.6%estimated ± 3.4 pp, low confidence
55Mercury 386.6%estimated ± 3.4 pp, low confidence
56Ornith-1.5-35B-A3B86.6%estimated ± 3.4 pp, medium confidence
57Qwen3.8-27B86.6%estimated ± 3.4 pp, medium confidence
58Solar Pro 486.5%estimated ± 3.4 pp, medium confidence
59Nemotron 3.5 Bolt 74B-A9B BF1686.5%estimated ± 3.4 pp, medium confidence
60Grok 4.2086.4%estimated ± 3.4 pp, medium confidence
61OrcaCyber-Nano-1.086.3%estimated ± 3.4 pp, medium confidence
62DeepSeek V4 Flash 073186.2%estimated ± 3.4 pp, medium confidence
63Inkling86.2%estimated ± 3.4 pp, medium confidence
64Hy3 Preview85.9%estimated ± 3.4 pp, medium confidence
65Nemotron 3 Ultra85.8%estimated ± 3.4 pp, medium confidence
66GLM-585.8%measured
67Step 3.7 Flash85.8%estimated ± 4.0 pp, medium confidence
68Inkling 2 Mini85.7%estimated ± 3.4 pp, medium confidence
69Nemotron 3.5 Bolt 74B-A9B85.6%estimated ± 3.4 pp, medium confidence
70Ornith-1.5-9B85.6%estimated ± 3.4 pp, medium confidence
71Solar Open 285.6%estimated ± 3.4 pp, medium confidence
72GLM-5.185.6%estimated ± 3.4 pp, medium confidence
73Ternary Bonsai 2 27B85.4%estimated ± 3.4 pp, medium confidence
74A.X K285.3%estimated ± 3.4 pp, medium confidence
75Ripple-65B Preview85.2%estimated ± 3.4 pp, medium confidence
76Ling 3.0 Flash85.1%estimated ± 3.4 pp, medium confidence
77MAI-Thinking-184.8%estimated ± 3.4 pp, medium confidence
78Ling 3.0 Flash FP884.8%estimated ± 3.4 pp, medium confidence
79Grok 484.6%estimated ± 4.0 pp, medium confidence
80Ling 3.0 Flash VL84.6%estimated ± 4.0 pp, medium confidence
81K-EXAONE 2.084.1%estimated ± 3.4 pp, medium confidence
82GLM-4.783.5%estimated ± 4.0 pp, medium confidence
83Qwen3.5 397B83.5%estimated ± 4.0 pp, medium confidence
84Trinity-Large-Thinking83.4%measured
85Mercury 2.582.8%estimated ± 3.4 pp, medium confidence
86Gemma 4 12B82.8%estimated ± 3.4 pp, medium confidence
87Sarvam 105B82.7%estimated ± 3.4 pp, medium confidence
88DeepSeek-R182.4%estimated ± 4.0 pp, medium confidence
89Gemini 2.5 Flash82.4%estimated ± 4.0 pp, medium confidence
90Gemini 3 Flash82.4%estimated ± 4.0 pp, medium confidence
91Gemma 4 31B82.4%estimated ± 4.0 pp, medium confidence
92GPT-OSS 20B82.4%estimated ± 4.0 pp, medium confidence
93Spark-X2.6-4B82.1%estimated ± 3.4 pp, medium confidence
94Nemotron 3.5 Lightning 30B A3B NVFP481.4%estimated ± 3.4 pp, medium confidence
95Claude 4 Sonnet81.3%estimated ± 4.0 pp, medium confidence
96GPT-OSS 120B81.3%estimated ± 4.0 pp, medium confidence
97K-Exaone81.3%estimated ± 4.0 pp, medium confidence
98MiMo-V2-Omni81.3%estimated ± 4.0 pp, medium confidence
99o381.3%estimated ± 4.0 pp, medium confidence
100Qwen3.6-27B81.3%estimated ± 4.0 pp, medium confidence
101LFM2.5-24B-A2B80.9%estimated ± 3.4 pp, medium confidence
102MiniMax M2.780.8%measured
103Claude 3.7 Sonnet Thinking80.5%estimated ± 4.0 pp, medium confidence
104Claude Sonnet 4.680.5%estimated ± 4.0 pp, medium confidence
105DeepSeek V3.280.5%estimated ± 4.0 pp, medium confidence
106Qwen3.5-27B80.5%estimated ± 4.0 pp, medium confidence
107Qwen3.5-35B-A3B80.5%estimated ± 4.0 pp, medium confidence
108Nemotron 3 Nano Omni 30B A3B80.0%estimated ± 3.4 pp, medium confidence
109ZAYA1-8B79.4%estimated ± 3.4 pp, medium confidence
110Gemma 4 E4B79.4%estimated ± 4.0 pp, medium confidence
111GLM-5V-Turbo79.4%estimated ± 4.0 pp, medium confidence
112o4-mini79.4%estimated ± 4.0 pp, medium confidence
113Qwen3.5-122B-A10B79.4%estimated ± 4.0 pp, medium confidence
114MiniCPM5-2B79.1%estimated ± 3.4 pp, medium confidence
115LongCat-Flash-Lite-Sparse78.7%estimated ± 3.4 pp, medium confidence
116Claude Opus 4.578.3%estimated ± 4.0 pp, low confidence
117Command A+78.3%estimated ± 4.0 pp, low confidence
118GLM-5-Turbo78.3%estimated ± 4.0 pp, low confidence
119Granite 4.2 30B78.3%estimated ± 4.0 pp, low confidence
120Granite 4.2 8B78.3%estimated ± 4.0 pp, low confidence
121Magistral Medium78.3%estimated ± 4.0 pp, low confidence
122MiMo-V2-Pro78.3%estimated ± 4.0 pp, low confidence
123Mistral Small 478.3%estimated ± 4.0 pp, low confidence
124North Mini Code78.3%estimated ± 4.0 pp, low confidence
125o178.3%estimated ± 4.0 pp, low confidence
126Qwen3.6-35B-A3B78.3%estimated ± 4.0 pp, low confidence
127Sarvam 30B77.3%estimated ± 3.4 pp, medium confidence
128Celeris-177.1%estimated ± 4.0 pp, low confidence
129Claude 3.5 Haiku77.1%estimated ± 4.0 pp, low confidence
130Claude 3.7 Sonnet77.1%estimated ± 4.0 pp, low confidence
131Claude 3 Haiku77.1%estimated ± 4.0 pp, low confidence
132DeepSeek V377.1%estimated ± 4.0 pp, low confidence
133DeepSeek V3 032477.1%estimated ± 4.0 pp, low confidence
134DeepSeek V3.177.1%estimated ± 4.0 pp, low confidence
135Devstral Small77.1%estimated ± 4.0 pp, low confidence
136Exaone 4.0 1.2B77.1%estimated ± 4.0 pp, low confidence
137Exaone 4.0 32B77.1%estimated ± 4.0 pp, low confidence
138Gemini 2.0 Flash77.1%estimated ± 4.0 pp, low confidence
139Gemini 2.5 Flash-Lite77.1%estimated ± 4.0 pp, low confidence
140Gemini 3.5 Flash-Lite77.1%estimated ± 4.0 pp, low confidence
141Gemma 3 27B77.1%estimated ± 4.0 pp, low confidence
142Gemma 4 26B A4B77.1%estimated ± 4.0 pp, low confidence
143Gemma 4 E2B77.1%estimated ± 4.0 pp, low confidence
144GLM-4.5-Air77.1%estimated ± 4.0 pp, low confidence
145GLM-4.677.1%estimated ± 4.0 pp, low confidence
146GPT-4.177.1%estimated ± 4.0 pp, low confidence
147GPT-4.1 mini77.1%estimated ± 4.0 pp, low confidence
148GPT-4.1 nano77.1%estimated ± 4.0 pp, low confidence
149GPT-4o77.1%estimated ± 4.0 pp, low confidence
150Granite-4.0-350M77.1%estimated ± 4.0 pp, low confidence
151Granite-4.0-H-1B77.1%estimated ± 4.0 pp, low confidence
152Granite-4.0-H-350M77.1%estimated ± 4.0 pp, low confidence
153Granite 4.2 3B77.1%estimated ± 4.0 pp, low confidence
154Grok 4.1 Fast77.1%estimated ± 4.0 pp, low confidence
155Grok Code Fast 177.1%estimated ± 4.0 pp, low confidence
156Kimi K277.1%estimated ± 4.0 pp, low confidence
157LFM2.5-2.6B77.1%estimated ± 4.0 pp, low confidence
158LFM2.5-8B-A1B77.1%estimated ± 4.0 pp, low confidence
159LFM2.5-VL-1.6B-Extract77.1%estimated ± 4.0 pp, low confidence
160Ling 2.6 Flash77.1%estimated ± 4.0 pp, low confidence
161Ling 3.0 Tiny77.1%estimated ± 4.0 pp, low confidence
162Llama 3.1 405B77.1%estimated ± 4.0 pp, low confidence
163Llama 3.1 70B77.1%estimated ± 4.0 pp, low confidence
164Llama 3.3 70B77.1%estimated ± 4.0 pp, low confidence
165Llama 4 Maverick77.1%estimated ± 4.0 pp, low confidence
166Llama 4 Scout77.1%estimated ± 4.0 pp, low confidence
167Magistral Small77.1%estimated ± 4.0 pp, low confidence
168MiMo-V2-Flash77.1%estimated ± 4.0 pp, low confidence
169Mistral Large 277.1%estimated ± 4.0 pp, low confidence
170Mistral Large 377.1%estimated ± 4.0 pp, low confidence
171Mistral Medium 377.1%estimated ± 4.0 pp, low confidence
172Mistral Medium 3.5 128B77.1%estimated ± 4.0 pp, low confidence
173Mistral Small 3.177.1%estimated ± 4.0 pp, low confidence
174Nemotron 3 Nano 30B77.1%estimated ± 4.0 pp, low confidence
175Nemotron Ultra 253B77.1%estimated ± 4.0 pp, low confidence
176Nova Pro77.1%estimated ± 4.0 pp, low confidence
177Phi-477.1%estimated ± 4.0 pp, low confidence
178Phi-4-reasoning77.1%estimated ± 4.0 pp, low confidence
179Qwen3.5-4B77.1%estimated ± 4.0 pp, low confidence
180Qwen3 Max77.1%estimated ± 4.0 pp, low confidence
181Qwen3-Omni-30B-A3B-Instruct77.1%estimated ± 4.0 pp, low confidence
182Qwen3-Omni-30B-A3B-Thinking77.1%estimated ± 4.0 pp, low confidence
183Solar Pro 277.1%estimated ± 4.0 pp, low confidence
184Solar Pro 377.1%estimated ± 4.0 pp, low confidence
185Ultravox v0.6 Llama 3.3 70B77.1%estimated ± 4.0 pp, low confidence
186Mellum2.1-12B-A2.5B-Thinking76.4%estimated ± 3.4 pp, medium confidence
187LFM3-1.7B75.7%estimated ± 3.4 pp, medium confidence
188Trinity-Large-Preview75.2%measured
189Mellum2-12B-A2.5B-Thinking72.6%estimated ± 3.4 pp, low confidence
190ZAYA1-74B-Preview72.4%estimated ± 3.4 pp, low confidence
191LLaDA2.2-mini63.9%estimated ± 3.4 pp, low confidence
192Soofi S 30B-A3B63.1%estimated ± 3.4 pp, low confidence
193Mellum2-12B-A2.5B-Instruct61.2%estimated ± 3.4 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General