benchgap
Knowledge & reasoning

SuperGPQA leaderboard

As of 2026-11-21, the highest measured score on SuperGPQA is 95.0% by Claude Opus 4.6. 261 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5100.0%estimated ± 9.3 pp, low confidence
2Claude Fable 5.1100.0%estimated ± 7.2 pp, low confidence
3Claude Fable 5.2100.0%estimated ± 7.2 pp, low confidence
4Claude Mythos 5100.0%estimated ± 7.2 pp, low confidence
5Claude Opus 5100.0%estimated ± 7.2 pp, low confidence
6Claude Opus 5.5100.0%estimated ± 9.3 pp, low confidence
7Claude Sonnet 5.5100.0%estimated ± 9.3 pp, low confidence
8Gemini 3.7 Flash100.0%estimated ± 9.3 pp, low confidence
9Gemini 3 Pro100.0%estimated ± 9.3 pp, low confidence
10Gemini 4 Carbon100.0%estimated ± 9.3 pp, low confidence
11Gemini 4 Neon100.0%estimated ± 9.3 pp, low confidence
12GPT-6.1 Sol100.0%estimated ± 9.3 pp, low confidence
13GPT-6 Sol100.0%estimated ± 9.3 pp, low confidence
14Grok 4.5100.0%estimated ± 9.3 pp, low confidence
15Kimi K3.1 Pro100.0%estimated ± 7.2 pp, low confidence
16Muse Spark 1.1100.0%estimated ± 7.2 pp, low confidence
17Gemini 3.8 Flash99.9%estimated ± 9.3 pp, low confidence
18Gemini 4 Hydrogen99.9%estimated ± 9.3 pp, low confidence
19Gemini 3.6 Flash99.1%estimated ± 9.3 pp, low confidence
20Hy497.8%estimated ± 7.2 pp, low confidence
21GPT-5.4 Pro97.4%estimated ± 7.2 pp, low confidence
22Claude Opus 4.896.7%estimated ± 7.2 pp, low confidence
23Gemini 4 Argon96.3%estimated ± 9.3 pp, low confidence
24Claude Sonnet 596.2%estimated ± 7.2 pp, low confidence
25Grok 4.696.2%estimated ± 9.3 pp, low confidence
26GPT-5.5 Pro96.0%estimated ± 7.2 pp, low confidence
27GPT-6.1 Terra95.1%estimated ± 9.3 pp, low confidence
28Claude Opus 4.695.0%measured
29Claude Sonnet 4.695.0%measured
30Apodex 1.195.0%estimated ± 7.2 pp, low confidence
31Kimi K394.9%estimated ± 7.2 pp, low confidence
32Hy4 preview94.3%estimated ± 7.2 pp, low confidence
33Hy4 flash94.1%estimated ± 7.2 pp, low confidence
34Claude Opus 4.6 (Adaptive)94.1%estimated ± 9.3 pp, low confidence
35Claude Opus 4.793.7%estimated ± 7.2 pp, low confidence
36Claude Opus 4.5 Thinking93.4%estimated ± 9.3 pp, low confidence
37Inkling 2 Small92.3%estimated ± 7.2 pp, low confidence
38Gemini 3 Flash91.9%estimated ± 9.3 pp, medium confidence
39dots3-note Preview91.7%estimated ± 7.2 pp, medium confidence
40GPT-5.591.3%estimated ± 7.2 pp, medium confidence
41GPT-5.491.2%estimated ± 7.2 pp, medium confidence
42Muse Spark 1.291.2%estimated ± 9.3 pp, medium confidence
43Qwen4 Max90.8%estimated ± 7.2 pp, medium confidence
44MiniMax M3.1 Flash90.6%estimated ± 9.3 pp, medium confidence
45MiniMax M390.3%estimated ± 9.3 pp, medium confidence
46Muse Spark89.6%estimated ± 7.2 pp, medium confidence
47Muse Spark 1.387.8%estimated ± 9.3 pp, medium confidence
48MiMo-V2.5-Pro87.4%estimated ± 7.2 pp, medium confidence
49Inkling-Small87.2%estimated ± 7.2 pp, medium confidence
50Agents-A187.0%estimated ± 7.2 pp, medium confidence
51Grok 4.786.2%estimated ± 9.3 pp, medium confidence
52Step 5 Preview86.0%estimated ± 7.2 pp, medium confidence
53Inkling85.5%estimated ± 7.2 pp, medium confidence
54Muse Spark Open84.9%estimated ± 9.3 pp, medium confidence
55GPT-6.2 Astra84.7%estimated ± 8.5 pp, low confidence
56Ornith-1.5-397B84.2%estimated ± 7.2 pp, medium confidence
57GPT-6.2 Sol83.4%estimated ± 8.5 pp, low confidence
58GPT-6 Astra83.4%estimated ± 8.5 pp, low confidence
59Qwen3.8 Max83.2%estimated ± 7.2 pp, medium confidence
60GPT-5.2-Codex83.2%estimated ± 9.3 pp, medium confidence
61Gemini 3.1 Pro83.0%estimated ± 13.9 pp, low confidence
62Muse Wonder Full83.0%estimated ± 9.3 pp, medium confidence
63Sakana Fugu82.8%estimated ± 8.5 pp, low confidence
64Sakana Fugu-Ultra82.8%estimated ± 8.5 pp, low confidence
65DeepSeek V4 Pro82.4%estimated ± 7.2 pp, medium confidence
66DeepSeek V4.1 Flash 101782.3%estimated ± 7.2 pp, medium confidence
67Grok 482.1%estimated ± 9.3 pp, medium confidence
68GPT-581.8%estimated ± 9.3 pp, medium confidence
69GPT-5.6 Sol81.6%estimated ± 8.5 pp, low confidence
70GPT-5.4 mini81.3%estimated ± 7.2 pp, medium confidence
71GPT-5.1-Codex81.1%estimated ± 9.3 pp, medium confidence
72GPT-5.1-Codex-Max81.1%estimated ± 9.3 pp, medium confidence
73GPT-6 Luna80.5%estimated ± 9.3 pp, medium confidence
74GLM-5.280.3%estimated ± 7.2 pp, medium confidence
75Pareto 26.10 Preview80.2%estimated ± 13.9 pp, low confidence
76Gemini 3.5 Flash80.0%estimated ± 7.2 pp, medium confidence
77GPT-5.6 Terra79.3%estimated ± 8.5 pp, low confidence
78GPT-5.278.6%estimated ± 8.5 pp, low confidence
79GPT-5.6 Luna78.5%estimated ± 8.5 pp, low confidence
80Muse Wonder77.8%estimated ± 9.3 pp, medium confidence
81GPT-5.4 nano77.7%estimated ± 7.2 pp, medium confidence
82Ling 3.0 Tiny77.7%estimated ± 9.3 pp, medium confidence
83Ling 3.1 Flash77.5%estimated ± 7.2 pp, medium confidence
84GPT-5.177.4%estimated ± 9.3 pp, medium confidence
85Particle-177.0%estimated ± 13.9 pp, low confidence
86Beam 1.0 501B-A23B FP876.9%estimated ± 13.9 pp, low confidence
87Qwen3.8-Omni-Flash76.6%estimated ± 7.2 pp, medium confidence
88Qwen3.8-Flash-Next76.0%estimated ± 7.2 pp, medium confidence
89Beam 1.0 501B-A23B NVFP475.3%estimated ± 13.9 pp, low confidence
90Grok 4.375.2%estimated ± 7.2 pp, medium confidence
91Interfaze Beta75.2%estimated ± 8.5 pp, low confidence
92Mercury 375.1%estimated ± 13.9 pp, low confidence
93DeepSeek V4 Flash 073175.0%estimated ± 7.2 pp, medium confidence
94Kimi K2.674.9%estimated ± 7.2 pp, medium confidence
95Grok 4.2073.9%estimated ± 13.9 pp, low confidence
96Qwen 3.6 Max (preview)73.9%measured
97Qwen3.7 Max73.6%measured
98OrcaCyber-Nano-1.073.6%estimated ± 13.9 pp, low confidence
99MiMo-V2.6-Pro73.2%estimated ± 9.3 pp, medium confidence
100Claude Haiku 5.573.1%estimated ± 9.3 pp, medium confidence
101Qwen3.6 Plus71.6%measured
102GLM-5.171.4%estimated ± 7.2 pp, medium confidence
103Qwen3.7 Plus71.4%measured
104Qwen3.8-27B71.2%estimated ± 7.2 pp, medium confidence
105Seed 2.1 Pro70.8%measured
106Claude Opus 4.570.6%measured
107Inkling 2 Mini70.5%estimated ± 8.5 pp, low confidence
108Nemotron 3.5 Bolt 74B-A9B70.5%estimated ± 13.9 pp, low confidence
109Qwen3.5 397B70.4%measured
110GLM-5.370.3%estimated ± 9.3 pp, medium confidence
111Solar Open 270.2%estimated ± 12.4 pp, low confidence
112Nemotron 3.5 Super VL 120B A12B69.7%estimated ± 7.2 pp, medium confidence
113Nemotron 3.5 Super VL 120B A12B NVFP469.5%estimated ± 7.2 pp, medium confidence
114Kimi K2.569.2%measured
115Beam 1.0 501B-A23B68.9%estimated ± 9.3 pp, medium confidence
116MiniMax M2.768.6%estimated ± 7.2 pp, medium confidence
117A.X K268.4%estimated ± 7.2 pp, medium confidence
118Qwen3.8 Max Preview68.0%estimated ± 9.3 pp, medium confidence
119GLM-5V-Turbo67.9%estimated ± 9.3 pp, medium confidence
120Seed 2.1 Turbo67.4%measured
121Nemotron 3 Ultra67.4%estimated ± 7.2 pp, medium confidence
122Ripple-65B Preview67.4%estimated ± 12.4 pp, low confidence
123GLM-5-Turbo67.3%estimated ± 9.3 pp, medium confidence
124Gemma 4 31B67.2%estimated ± 7.2 pp, medium confidence
125Apodex 1.1 Mini67.2%estimated ± 9.3 pp, medium confidence
126Qwen3.5-122B-A10B67.1%measured
127MAI-Thinking-167.1%estimated ± 8.5 pp, low confidence
128Muse Glimmer 30B67.1%estimated ± 9.3 pp, medium confidence
129Ling 3.0 Flash FP866.8%estimated ± 8.5 pp, low confidence
130GLM-566.8%measured
131DeepSeek-R1-052866.7%estimated ± 4.1 pp, high confidence
132MiMo-V2.6-Flash66.6%estimated ± 9.3 pp, medium confidence
133MiMo-V2-Flash66.4%estimated ± 8.5 pp, low confidence
134MiMo-V2-Pro66.4%estimated ± 9.3 pp, medium confidence
135Ornith-1.5-35B-A3B66.3%estimated ± 7.2 pp, medium confidence
136Gemini 3.5 Flash-Lite66.3%estimated ± 9.3 pp, medium confidence
137Hy3 Preview66.2%estimated ± 7.2 pp, medium confidence
138Gemini 2.5 Flash66.2%estimated ± 9.3 pp, medium confidence
139Mistral Large 466.1%estimated ± 9.3 pp, medium confidence
140Step 3.7 Flash66.0%estimated ± 9.3 pp, medium confidence
141Qwen3.6-27B66.0%measured
142Claude Sonnet 4.566.0%estimated ± 8.5 pp, low confidence
143Grok 4.1 Fast (Reasoning)65.8%estimated ± 9.3 pp, medium confidence
144Mistral Large 365.8%estimated ± 9.3 pp, medium confidence
145Qwen3.5-27B65.6%measured
146GLM-4.765.6%estimated ± 7.2 pp, medium confidence
147Qwen3 Max65.6%estimated ± 9.3 pp, medium confidence
148DeepSeek V3 032465.5%estimated ± 9.3 pp, medium confidence
149DeepSeek V3.265.5%estimated ± 9.3 pp, medium confidence
150Grok Code Fast 165.3%estimated ± 9.3 pp, medium confidence
151Mistral Medium 3.5 128B65.2%estimated ± 9.3 pp, medium confidence
152Grok 4 Fast (Reasoning)65.2%estimated ± 9.3 pp, medium confidence
153Claude 4 Sonnet65.1%estimated ± 9.3 pp, medium confidence
154Mercury 2.565.0%estimated ± 9.3 pp, medium confidence
155Nemotron 3.5 Bolt 74B-A9B BF1665.0%estimated ± 9.3 pp, medium confidence
156GLM-4.664.9%estimated ± 9.3 pp, medium confidence
157Mistral Small 464.9%estimated ± 9.3 pp, medium confidence
158Muse Ember64.9%estimated ± 9.3 pp, medium confidence
159GPT-4o64.7%estimated ± 9.3 pp, medium confidence
160Mistral Large 264.7%estimated ± 9.3 pp, medium confidence
161Qwen3.6-35B-A3B64.7%measured
162MiMo-V2-Omni64.7%estimated ± 9.3 pp, low confidence
163Solar Pro 364.6%estimated ± 9.3 pp, low confidence
164North Mini Code64.6%estimated ± 9.3 pp, low confidence
165Solar Pro 464.6%estimated ± 9.3 pp, low confidence
166Mistral Medium 364.6%estimated ± 9.3 pp, low confidence
167Claude 3 Haiku64.5%estimated ± 9.3 pp, low confidence
168Grok 4.1 Fast64.5%estimated ± 9.3 pp, low confidence
169Nova Pro64.5%estimated ± 9.3 pp, low confidence
170K-Exaone64.5%estimated ± 9.3 pp, low confidence
171GLM-4.5-Air64.5%estimated ± 9.3 pp, low confidence
172Solar Pro 264.4%estimated ± 9.3 pp, low confidence
173Qwen3 235B-A22B (Reasoning)64.4%estimated ± 2.0 pp, low confidence
174Mistral Small 3.164.4%estimated ± 9.3 pp, low confidence
175Devstral Small64.4%estimated ± 9.3 pp, low confidence
176GPT-4.1 nano64.4%estimated ± 9.3 pp, low confidence
177Claude 3.5 Haiku64.4%estimated ± 9.3 pp, low confidence
178Gemma 3 27B64.4%estimated ± 9.3 pp, low confidence
179Command A+64.4%estimated ± 9.3 pp, low confidence
180LFM2.5-8B-A1B64.4%estimated ± 9.3 pp, low confidence
181Granite 4.2 3B64.3%estimated ± 9.3 pp, low confidence
182Exaone 4.0 32B64.0%estimated ± 12.4 pp, low confidence
183Sarvam 105B63.9%estimated ± 12.4 pp, low confidence
184Ling 3.0 Flash63.6%estimated ± 7.2 pp, medium confidence
185Qwen3.5-35B-A3B63.4%measured
186Qwen3 235B 250762.6%measured
187o3-pro62.5%estimated ± 10.1 pp, low confidence
188QwQ-32B62.4%estimated ± 2.0 pp, low confidence
189Grok 3 Mini61.7%estimated ± 8.5 pp, low confidence
190Claude 4.1 Opus61.7%estimated ± 10.1 pp, low confidence
191o1-preview61.7%estimated ± 10.1 pp, low confidence
192Claude 3 Opus61.7%estimated ± 10.1 pp, low confidence
193DeepSeek R1 Distill Qwen 32B61.7%estimated ± 10.1 pp, low confidence
194Gemini 1.5 Pro61.7%estimated ± 10.1 pp, low confidence
195Gemini 1.5 Flash61.7%estimated ± 10.1 pp, low confidence
196GPT-4 Turbo61.7%estimated ± 10.1 pp, low confidence
197Qwen2.5 Coder 32B Instruct61.7%estimated ± 10.1 pp, low confidence
198GPT-4o mini61.7%estimated ± 10.1 pp, low confidence
199DeepSeek-V2.561.7%estimated ± 10.1 pp, low confidence
200Gemini 1.0 Pro61.7%estimated ± 10.1 pp, low confidence
201Sarvam 30B61.6%estimated ± 12.4 pp, low confidence
202Ornith-1.5-9B61.3%estimated ± 7.2 pp, medium confidence
203Claude 4 Opus Thinking60.7%estimated ± 8.5 pp, low confidence
204LongCat-Flash-Lite-Sparse60.6%estimated ± 12.4 pp, low confidence
205Qwen3 32B (Reasoning)60.5%estimated ± 2.0 pp, low confidence
206Gemini 2.5 Pro59.9%estimated ± 7.2 pp, medium confidence
207o1-pro59.9%estimated ± 8.5 pp, low confidence
208Gemma 4 12B59.7%estimated ± 8.5 pp, low confidence
209K-EXAONE 2.059.5%estimated ± 7.2 pp, medium confidence
210Nemotron 3 Super 100B59.4%estimated ± 7.2 pp, medium confidence
211Claude 3.7 Sonnet Thinking58.9%estimated ± 8.5 pp, low confidence
212Gemma 4 26B A4B58.4%estimated ± 7.2 pp, medium confidence
213Qwen3 235B-A22B58.3%estimated ± 2.0 pp, low confidence
214Ternary Bonsai 2 27B58.3%estimated ± 4.1 pp, high confidence
215o4-mini58.1%estimated ± 8.5 pp, low confidence
216o3-mini57.6%estimated ± 8.5 pp, low confidence
217Kimi K257.2%measured
218Celeris-156.3%estimated ± 12.4 pp, low confidence
219GPT-OSS 120B56.3%estimated ± 7.2 pp, medium confidence
220Nemotron Ultra 253B56.1%estimated ± 8.5 pp, low confidence
221MiniMax-Text-0156.1%estimated ± 12.4 pp, low confidence
222Grok 255.8%estimated ± 12.4 pp, low confidence
223o155.7%estimated ± 8.5 pp, low confidence
224Mellum2.1-12B-A2.5B-Thinking55.6%estimated ± 4.1 pp, high confidence
225Nemotron 3.5 Lightning 30B A3B NVFP455.6%estimated ± 8.5 pp, low confidence
226Grok 3 [Beta]55.4%estimated ± 8.5 pp, low confidence
227GLM-4.7-Flash55.1%estimated ± 8.5 pp, low confidence
228Claude 4 Opus54.8%estimated ± 8.5 pp, low confidence
229DeepSeek V3.154.8%estimated ± 8.5 pp, low confidence
230Trinity-Large-Thinking53.0%estimated ± 13.9 pp, low confidence
231GPT-OSS 20B52.5%estimated ± 7.2 pp, medium confidence
232Mellum2-12B-A2.5B-Thinking52.4%estimated ± 4.1 pp, high confidence
233Nemotron 3 Nano 30B52.2%estimated ± 7.2 pp, medium confidence
234Nemotron 3 Nano Omni 30B A3B51.7%estimated ± 8.5 pp, low confidence
235Grok 2 mini51.6%estimated ± 12.4 pp, low confidence
236Qwen3 32B51.3%estimated ± 4.1 pp, high confidence
237DeepSeek-R150.9%estimated ± 8.5 pp, low confidence
238ZAYA1-8B50.4%estimated ± 8.5 pp, low confidence
239Magistral Medium50.2%estimated ± 8.5 pp, low confidence
240Spark-X2.6-4B49.3%estimated ± 8.5 pp, low confidence
241LFM2.5-24B-A2B49.2%estimated ± 7.2 pp, medium confidence
242LFM3-1.7B49.0%estimated ± 7.2 pp, medium confidence
243GPT-4.548.9%estimated ± 8.5 pp, low confidence
244Phi-4-reasoning-plus48.3%estimated ± 8.5 pp, low confidence
245Llama 3.3 70B48.0%estimated ± 12.4 pp, low confidence
246Qwen2.5-VL-32B47.9%estimated ± 12.4 pp, low confidence
247Magistral Small47.7%estimated ± 8.5 pp, low confidence
248Claude 3.7 Sonnet47.5%estimated ± 8.5 pp, low confidence
249ZAYA1-74B-Preview47.1%estimated ± 12.4 pp, low confidence
250Granite 4.2 30B46.1%estimated ± 8.5 pp, low confidence
251GPT-4.146.0%estimated ± 8.5 pp, low confidence
252Phi-4-reasoning45.6%estimated ± 8.5 pp, low confidence
253Llama 3.1 70B45.3%estimated ± 12.4 pp, low confidence
254DeepSeek R1 Distill Llama 70B45.1%estimated ± 8.5 pp, low confidence
255Gemini 2.0 Flash45.1%estimated ± 8.5 pp, low confidence
256Gemini 2.5 Flash-Lite44.6%estimated ± 8.5 pp, low confidence
257GPT-4.1 mini44.3%estimated ± 8.5 pp, low confidence
258Granite 4.2 8B44.3%estimated ± 8.5 pp, low confidence
259DeepSeek-R1-0528-Qwen3-8B42.2%estimated ± 8.5 pp, low confidence
260Qwen2.5-Max41.6%estimated ± 8.5 pp, low confidence
261o1-mini41.6%estimated ± 8.5 pp, low confidence
262Claude 3.5 Sonnet41.3%estimated ± 8.5 pp, low confidence
263DeepSeek R1 Distill Qwen 14B41.1%estimated ± 8.5 pp, low confidence
264DeepSeek V341.1%estimated ± 8.5 pp, low confidence
265Ling 2.6 Flash41.0%estimated ± 8.5 pp, low confidence
266Gemma 4 E4B40.8%estimated ± 8.5 pp, low confidence
267MiniCPM5-2B40.8%measured
268Gemma 4 E2B38.6%estimated ± 12.4 pp, low confidence
269Mellum2-12B-A2.5B-Instruct36.0%estimated ± 4.1 pp, high confidence
270Trinity-Large-Preview34.8%estimated ± 13.9 pp, low confidence
271Soofi S 30B-A3B30.6%estimated ± 12.4 pp, low confidence
272Llama 3.1 8B27.9%estimated ± 12.4 pp, low confidence
273Phi-3.5-mini27.1%estimated ± 12.4 pp, low confidence
274LLaDA2.2-mini23.7%estimated ± 13.9 pp, low confidence
275MiniCPM5-1B23.1%measured
276Granite-4.0-1B16.7%estimated ± 12.4 pp, low confidence
277Granite-4.0-H-1B15.9%estimated ± 12.4 pp, low confidence
278LFM2.5-230M7.6%estimated ± 12.4 pp, low confidence
279LFM2.5-350M7.4%estimated ± 12.4 pp, low confidence
280LFM2.5-VL-450M7.0%estimated ± 12.4 pp, low confidence
281Granite-4.0-H-350M4.1%estimated ± 12.4 pp, low confidence
282Granite-4.0-350M2.8%estimated ± 12.4 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General