benchgap
Knowledge & reasoning

Vals GPQA Diamond leaderboard

As of 2026-11-21, the highest measured score on Vals GPQA Diamond is 95.5% by Gemini 3.1 Pro. 223 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Sonnet 5.598.8%estimated ± 1.2 pp, low confidence
2GPT-6.2 Astra98.8%estimated ± 1.2 pp, low confidence
3GPT-6.2 Sol98.8%estimated ± 1.2 pp, low confidence
4Claude Opus 5.598.8%estimated ± 1.2 pp, low confidence
5Ling 3.1 Flash98.8%estimated ± 1.2 pp, low confidence
6GPT-6 Astra98.8%estimated ± 1.2 pp, low confidence
7GPT-6.1 Sol98.8%estimated ± 1.2 pp, low confidence
8GPT-6.1 Terra98.7%estimated ± 1.2 pp, low confidence
9Muse Wonder Full98.6%estimated ± 1.2 pp, low confidence
10Claude Fable 5.296.5%estimated ± 2.9 pp, medium confidence
11GPT-6 Luna96.3%estimated ± 1.2 pp, low confidence
12GPT-6 Sol96.3%estimated ± 1.2 pp, low confidence
13Gemini 3.1 Pro95.5%measured
14Muse Wonder95.3%estimated ± 1.2 pp, medium confidence
15GPT-5.6 Sol95.2%measured
16Grok 4.694.7%measured
17Claude Mythos 594.4%estimated ± 2.9 pp, high confidence
18Gemini 3.8 Flash94.4%measured
19Sakana Fugu94.3%estimated ± 2.1 pp, medium confidence
20Sakana Fugu-Ultra94.3%estimated ± 2.1 pp, medium confidence
21Gemini 4 Argon94.1%estimated ± 1.4 pp, medium confidence
22Gemini 3.7 Flash93.9%measured
23Gemini 4 Carbon93.8%estimated ± 2.3 pp, high confidence
24Qwen3.8 Max93.7%measured
25Qwen4 Max93.5%estimated ± 1.2 pp, medium confidence
26Claude Fable 5.193.4%measured
27Claude Opus 593.4%measured
28Gemini 3.6 Flash93.4%measured
29GPT-5.5 Pro93.2%estimated ± 2.3 pp, high confidence
30Claude Fable 593.2%measured
31GPT-5.593.2%measured
32GPT-5.4 Pro93.2%estimated ± 2.3 pp, high confidence
33Kimi K3.1 Pro93.1%estimated ± 2.1 pp, high confidence
34Claude Opus 4.693.1%estimated ± 1.4 pp, high confidence
35Gemini 4 Neon93.1%estimated ± 5.7 pp, low confidence
36dots3-note Preview93.0%estimated ± 2.3 pp, high confidence
37Claude Haiku 5.593.0%estimated ± 1.4 pp, high confidence
38Grok 4.592.9%measured
39Kimi K392.9%measured
40Gemini 3.5 Flash92.7%measured
41MiniMax M392.7%measured
42Gemini 4 Hydrogen92.7%estimated ± 5.7 pp, medium confidence
43Muse Spark 1.392.6%estimated ± 3.2 pp, high confidence
44Step 5 Preview92.5%estimated ± 2.1 pp, high confidence
45Inkling 2 Small92.4%estimated ± 1.4 pp, high confidence
46Claude Opus 4.892.4%measured
47DeepSeek V4 Pro92.4%measured
48Pareto 26.992.3%estimated ± 2.9 pp, high confidence
49MiMo-V2.6-Pro92.1%estimated ± 5.7 pp, medium confidence
50Qwen3.8 Max Preview91.9%estimated ± 3.2 pp, high confidence
51Ornith-1.5-397B91.8%estimated ± 2.1 pp, high confidence
52Gemini 3 Pro91.7%measured
53GPT-5.291.7%measured
54GPT-5.491.7%measured
55GPT-5.6 Luna91.7%measured
56Sage-A1.591.7%estimated ± 7.0 pp, medium confidence
57ERNIE 5.5 Flash Preview91.6%estimated ± 2.9 pp, high confidence
58Pareto 26.10 Preview91.5%estimated ± 2.1 pp, high confidence
59Grok 4.391.4%measured
60Hy4 preview91.4%estimated ± 2.1 pp, high confidence
61Grok 4.791.3%estimated ± 1.2 pp, medium confidence
62Ling 3.0 Flash Sante91.3%estimated ± 1.2 pp, medium confidence
63Muse Spark 1.191.2%measured
64GPT-5.6 Terra90.9%measured
65Qwen3.8-Flash-Next90.8%estimated ± 2.1 pp, high confidence
66MiniMax M3.1 Flash90.8%estimated ± 2.9 pp, high confidence
67Muse Ember90.7%estimated ± 5.7 pp, medium confidence
68GPT-5.3 Codex90.6%estimated ± 3.2 pp, high confidence
69Claude Opus 4.790.2%measured
70Qwen3.7 Max90.2%measured
71Qwen3.8-Omni-Flash90.1%estimated ± 2.1 pp, high confidence
72DeepSeek V4.1 Flash 101790.0%estimated ± 2.1 pp, high confidence
73DeepSeek V4 Flash 073189.9%measured
74Beam 1.0 501B-A23B89.8%estimated ± 2.1 pp, high confidence
75Claude Opus 4.6 (Adaptive)89.6%measured
76Muse Spark89.6%measured
77Particle-189.6%estimated ± 2.1 pp, high confidence
78Muse Spark 1.289.5%estimated ± 3.2 pp, high confidence
79Qwen3.7 Plus89.5%estimated ± 2.1 pp, high confidence
80Beam 1.0 501B-A23B FP889.5%estimated ± 2.1 pp, high confidence
81Kimi K2.689.1%measured
82Interfaze Beta89.1%estimated ± 2.1 pp, high confidence
83Nemotron 3.5 Super VL 120B A12B NVFP489.1%estimated ± 2.1 pp, high confidence
84GPT-5.2-Codex89.0%estimated ± 3.2 pp, high confidence
85Claude Sonnet 588.9%measured
86Qwen3.8-27B88.9%measured
87Hy388.8%estimated ± 3.2 pp, high confidence
88Nemotron 3.5 Super VL 120B A12B88.7%estimated ± 2.1 pp, high confidence
89Kimi K2.7 Code88.7%estimated ± 3.2 pp, high confidence
90Grok 4.2088.6%measured
91Beam 1.0 501B-A23B NVFP488.5%estimated ± 2.1 pp, high confidence
92Mercury 388.4%estimated ± 2.1 pp, high confidence
93Ornith-1.5-35B-A3B88.4%estimated ± 2.1 pp, high confidence
94Qwen3.5 397B88.4%estimated ± 3.2 pp, high confidence
95Solar Pro 488.2%estimated ± 2.1 pp, high confidence
96GLM-5.388.1%measured
97Grok 488.1%measured
98Muse Spark Open88.1%estimated ± 3.2 pp, high confidence
99Nemotron 3.5 Bolt 74B-A9B BF1688.0%estimated ± 2.1 pp, high confidence
100MiMo-V2.6-Flash88.0%estimated ± 5.7 pp, medium confidence
101Gemini 3 Flash87.9%measured
102Qwen 3.6 Max (preview)87.9%estimated ± 3.2 pp, high confidence
103Mistral Large 487.7%estimated ± 5.7 pp, medium confidence
104OrcaCyber-Nano-1.087.5%estimated ± 2.1 pp, high confidence
105Qwen3.6 Plus87.4%measured
106Inkling87.1%measured
107Claude 4 Opus Thinking87.0%estimated ± 3.1 pp, medium confidence
108Kimi K2.586.9%estimated ± 2.1 pp, high confidence
109GPT-5.186.6%measured
110MiniMax M2.786.6%measured
111Hy3 Preview86.5%estimated ± 2.1 pp, high confidence
112GLM-5.3-Flash86.4%measured
113MiMo-V2-Pro86.1%estimated ± 3.2 pp, high confidence
114Nemotron 3 Ultra86.1%measured
115Claude Opus 4.5 Thinking85.9%measured
116Inkling 2 Mini85.9%estimated ± 2.1 pp, high confidence
117Nemotron 3.5 Bolt 74B-A9B85.7%estimated ± 2.1 pp, high confidence
118Ornith-1.5-9B85.7%estimated ± 2.1 pp, high confidence
119Claude Sonnet 4.685.6%measured
120GLM-5.285.6%measured
121GPT-585.6%measured
122Solar Open 285.6%estimated ± 2.1 pp, high confidence
123Apodex 1.185.5%estimated ± 3.2 pp, high confidence
124Apodex 1.1 Mini85.5%estimated ± 3.2 pp, high confidence
125Grok 4 Fast (Reasoning)85.4%measured
126GLM-585.3%estimated ± 2.1 pp, high confidence
127Ling 3.0 Flash VL85.3%estimated ± 3.2 pp, high confidence
128Gemini 2.5 Pro85.1%estimated ± 3.1 pp, medium confidence
129GPT-5.1-Codex85.1%estimated ± 3.2 pp, high confidence
130GPT-5.1-Codex-Max85.1%estimated ± 3.2 pp, high confidence
131Ternary Bonsai 2 27B85.1%estimated ± 2.1 pp, high confidence
132A.X K284.9%estimated ± 2.1 pp, high confidence
133Qwen3.5-27B84.9%estimated ± 3.2 pp, high confidence
134Qwen3.5-122B-A10B84.8%estimated ± 3.2 pp, high confidence
135Ling 3.0 Flash84.8%measured
136Ripple-65B Preview84.5%estimated ± 2.1 pp, high confidence
137GLM-5.184.5%measured
138Grok 4.1 Fast (Reasoning)84.3%measured
139o384.1%measured
140GLM-5-Turbo83.8%estimated ± 3.2 pp, high confidence
141Gemini 3.5 Flash-Lite83.8%measured
142o3-pro83.6%estimated ± 3.2 pp, high confidence
143Qwen3.5-35B-A3B83.6%estimated ± 3.2 pp, high confidence
144Inkling-Small83.6%measured
145MAI-Thinking-183.6%estimated ± 2.1 pp, medium confidence
146Ling 3.0 Flash FP883.4%estimated ± 2.1 pp, medium confidence
147Qwen3.6-27B83.3%estimated ± 3.2 pp, high confidence
148Qwen3.6-35B-A3B83.2%estimated ± 3.2 pp, high confidence
149GPT-5.4 mini83.1%measured
150DeepSeek-R182.9%estimated ± 3.1 pp, low confidence
151Qwen3.5 Flash82.8%measured
152Muse Glimmer 30B82.6%estimated ± 3.2 pp, high confidence
153MiMo-V2.5-Pro82.6%measured
154Claude 4.1 Opus82.4%estimated ± 5.7 pp, medium confidence
155MiMo-V2-Omni81.9%estimated ± 3.2 pp, high confidence
156K-EXAONE 2.081.7%estimated ± 2.1 pp, medium confidence
157Claude Sonnet 4.5 Thinking81.6%measured
158MiMo-V2.581.6%measured
159Claude Sonnet 4.581.5%estimated ± 2.3 pp, high confidence
160Gemini 3.1 Flash-Lite81.1%measured
161GLM-4.780.1%measured
162GLM-5V-Turbo80.0%estimated ± 3.2 pp, high confidence
163Step 3.7 Flash80.0%estimated ± 3.2 pp, high confidence
164Claude Opus 4.579.5%measured
165Qwen3 Max79.5%measured
166DeepSeek-R1-052878.8%estimated ± 4.5 pp, high confidence
167Mercury 2.578.7%estimated ± 2.1 pp, medium confidence
168Gemma 4 12B78.5%estimated ± 2.1 pp, medium confidence
169Sarvam 105B78.4%estimated ± 2.1 pp, medium confidence
170Grok 3 Mini78.1%estimated ± 4.5 pp, high confidence
171GPT-5.4 nano77.5%measured
172K-Exaone77.4%estimated ± 3.2 pp, high confidence
173GPT-OSS 120B77.3%estimated ± 3.2 pp, high confidence
174Gemma 4 31B77.1%estimated ± 2.9 pp, medium confidence
175Nemotron 3 Super 100B77.1%estimated ± 4.5 pp, high confidence
176Spark-X2.6-4B77.0%estimated ± 2.1 pp, medium confidence
177o1-pro76.8%estimated ± 4.5 pp, high confidence
178Trinity-Large-Thinking76.3%estimated ± 2.1 pp, medium confidence
179Qwen3.5-4B76.2%estimated ± 3.2 pp, high confidence
180Mistral Small 476.0%estimated ± 3.2 pp, high confidence
181Nemotron 3.5 Lightning 30B A3B NVFP475.7%estimated ± 2.1 pp, medium confidence
182Kimi K275.7%estimated ± 3.2 pp, high confidence
183o1-preview75.6%estimated ± 3.2 pp, high confidence
184o3-mini75.5%measured
185Qwen3 235B 250775.3%estimated ± 4.5 pp, high confidence
186Claude 3.7 Sonnet Thinking75.3%measured
187Command A+75.2%estimated ± 3.2 pp, high confidence
188North Mini Code74.8%estimated ± 3.2 pp, high confidence
189GLM-4.674.5%measured
190o4-mini74.5%measured
191Trinity-Large-Preview74.3%estimated ± 3.2 pp, high confidence
192DeepSeek V3.274.2%estimated ± 3.2 pp, high confidence
193Grok 3 [Beta]73.3%estimated ± 4.5 pp, high confidence
194o173.2%measured
195GLM-4.7-Flash73.1%estimated ± 4.5 pp, high confidence
196DeepSeek V3.172.6%estimated ± 3.2 pp, high confidence
197Ling 3.0 Tiny72.5%estimated ± 3.2 pp, high confidence
198GLM-4.5-Air72.4%estimated ± 3.2 pp, high confidence
199Quasar 438B72.3%estimated ± 3.2 pp, high confidence
200LFM2.5-24B-A2B72.3%estimated ± 4.5 pp, high confidence
201Claude Haiku 4.572.2%measured
202GLM-4.572.2%measured
203Nemotron Ultra 253B71.9%estimated ± 3.2 pp, high confidence
204Grok Code Fast 171.8%estimated ± 3.2 pp, high confidence
205Qwen3-Omni-30B-A3B-Thinking71.7%estimated ± 3.2 pp, high confidence
206Solar Pro 371.5%estimated ± 3.2 pp, high confidence
207MiniCPM5-2B69.3%estimated ± 3.2 pp, high confidence
208Claude 4 Opus69.2%estimated ± 3.2 pp, high confidence
209Qwen3 235B-A22B (Reasoning)69.0%estimated ± 4.5 pp, high confidence
210ZAYA1-8B68.9%estimated ± 4.5 pp, high confidence
211Mistral Large 368.4%measured
212GPT-OSS 20B67.9%estimated ± 3.2 pp, high confidence
213GPT-4.1 mini67.9%measured
214Claude 4 Sonnet67.4%estimated ± 3.2 pp, high confidence
215Gemini 2.5 Flash67.4%estimated ± 3.2 pp, high confidence
216GPT-4.567.4%estimated ± 4.5 pp, high confidence
217Claude 3.7 Sonnet67.4%measured
218Magistral Medium67.0%estimated ± 3.2 pp, high confidence
219Phi-4-reasoning-plus66.8%estimated ± 4.5 pp, high confidence
220Qwen3 32B (Reasoning)66.3%estimated ± 4.5 pp, high confidence
221Llama 4 Maverick66.2%estimated ± 3.2 pp, high confidence
222GPT-4.165.4%measured
223Gemini 2.0 Flash65.2%measured
224DeepSeek V3 032464.6%estimated ± 3.2 pp, high confidence
225Granite 4.2 30B63.5%estimated ± 3.2 pp, high confidence
226Magistral Small63.2%estimated ± 3.2 pp, high confidence
227DeepSeek R1 Distill Llama 70B63.1%estimated ± 4.5 pp, high confidence
228Grok 4.1 Fast62.8%estimated ± 3.2 pp, high confidence
229Sarvam 30B62.4%estimated ± 3.2 pp, high confidence
230Celeris-162.2%estimated ± 3.2 pp, high confidence
231Granite 4.2 8B62.2%estimated ± 3.2 pp, high confidence
232Exaone 4.0 32B61.9%estimated ± 3.2 pp, high confidence
233LFM3-1.7B61.3%estimated ± 4.5 pp, high confidence
234Qwen3-Omni-30B-A3B-Instruct61.1%estimated ± 3.2 pp, high confidence
235Qwen3 235B-A22B60.9%estimated ± 4.5 pp, high confidence
236DeepSeek R1 Distill Qwen 32B60.6%estimated ± 3.2 pp, high confidence
237LongCat-Flash-Lite-Sparse59.8%estimated ± 8.2 pp, medium confidence
238o1-mini59.4%estimated ± 3.2 pp, high confidence
239Claude 3.5 Sonnet59.3%measured
240MiMo-V2-Flash59.3%measured
241DeepSeek-R1-0528-Qwen3-8B59.1%estimated ± 4.5 pp, high confidence
242Ling 2.6 Flash58.4%estimated ± 3.2 pp, high confidence
243QwQ-32B58.4%estimated ± 3.2 pp, high confidence
244Gemini 1.5 Pro58.3%measured
245Qwen2.5-Max58.1%estimated ± 4.5 pp, high confidence
246Llama 4 Scout57.8%estimated ± 3.2 pp, high confidence
247Gemma 4 26B A4B57.5%estimated ± 2.9 pp, medium confidence
248DeepSeek R1 Distill Qwen 14B57.1%estimated ± 4.5 pp, high confidence
249Mistral Medium 356.9%estimated ± 3.2 pp, high confidence
250Gemma 4 E4B56.7%estimated ± 3.2 pp, high confidence
251Phi-456.6%estimated ± 3.2 pp, high confidence
252Phi-4-reasoning56.6%estimated ± 3.2 pp, high confidence
253Mellum2-12B-A2.5B-Thinking55.6%estimated ± 4.5 pp, high confidence
254ZAYA1-74B-Preview55.3%estimated ± 4.5 pp, high confidence
255Solar Pro 255.2%estimated ± 3.2 pp, high confidence
256Laguna XS.255.1%measured
257Granite 4.2 3B55.0%estimated ± 3.2 pp, high confidence
258LFM2.5-2.6B54.9%estimated ± 3.2 pp, high confidence
259DeepSeek V354.8%estimated ± 3.2 pp, high confidence
260Grok 254.0%estimated ± 4.5 pp, high confidence
261GPT-4o53.8%measured
262Qwen3 32B52.6%estimated ± 4.5 pp, high confidence
263MiniMax-Text-0152.4%estimated ± 4.5 pp, high confidence
264GPT-4.1 nano50.8%measured
265Llama 3.1 405B50.6%estimated ± 3.2 pp, high confidence
266LFM2.5-8B-A1B50.4%estimated ± 3.2 pp, high confidence
267GPT-4 Turbo49.8%estimated ± 5.7 pp, medium confidence
268Grok 2 mini49.1%estimated ± 4.5 pp, high confidence
269Nova Pro49.0%estimated ± 3.2 pp, high confidence
270Llama 3.3 70B48.9%estimated ± 3.2 pp, high confidence
271Ultravox v0.6 Llama 3.3 70B48.9%estimated ± 3.2 pp, high confidence
272Gemini 2.0 Flash-Lite48.6%estimated ± 4.5 pp, high confidence
273Llama 3.1 70B48.2%estimated ± 3.2 pp, high confidence
274Claude 3 Opus48.0%estimated ± 3.2 pp, high confidence
275Mistral Large 247.7%estimated ± 3.2 pp, high confidence
276Gemini 2.5 Flash-Lite46.6%estimated ± 3.2 pp, high confidence
277Nemotron 3 Nano Omni 30B A3B46.1%estimated ± 3.2 pp, high confidence
278Gemini 1.5 Flash46.0%measured
279Mistral Small 3.144.6%estimated ± 3.2 pp, high confidence
280GPT-4o mini44.2%measured
281Qwen2.5-VL-32B44.1%estimated ± 4.5 pp, high confidence
282Mixtral 8x22B Instruct v0.143.6%estimated ± 8.2 pp, low confidence
283Gemma 4 E2B42.5%estimated ± 3.2 pp, high confidence
284Gemma 3 27B42.0%estimated ± 3.2 pp, high confidence
285Exaone 4.0 1.2B41.6%estimated ± 3.2 pp, high confidence
286Soofi S 30B-A3B41.5%estimated ± 4.5 pp, high confidence
287DeepSeek-V2.541.5%estimated ± 3.2 pp, high confidence
288Qwen2.5 Coder 32B Instruct40.9%estimated ± 3.2 pp, high confidence
289Devstral Small40.6%estimated ± 3.2 pp, high confidence
290Gemma 2 27B40.2%estimated ± 8.2 pp, low confidence
291Mellum2-12B-A2.5B-Instruct39.1%estimated ± 4.5 pp, medium confidence
292Nemotron 3 Nano 30B39.1%estimated ± 3.2 pp, medium confidence
293Claude 3.5 Haiku37.9%measured
294Claude 3 Haiku36.6%estimated ± 3.2 pp, medium confidence
295Mistral Medium 3.5 128B34.8%measured
296Phi-4 Multimodal Instruct30.7%estimated ± 3.2 pp, medium confidence
297LFM2.5-350M28.9%estimated ± 4.5 pp, medium confidence
298Phi-3.5-mini28.7%estimated ± 4.5 pp, medium confidence
299Granite-4.0-1B28.2%estimated ± 4.5 pp, medium confidence
300LFM2.5-VL-1.6B-Extract28.1%estimated ± 3.2 pp, medium confidence
301Laguna M.127.0%measured
302Gemini 1.0 Pro26.9%estimated ± 3.2 pp, medium confidence
303Llama 3.1 8B26.2%estimated ± 3.2 pp, medium confidence
304Granite-4.0-H-1B25.5%estimated ± 3.2 pp, medium confidence
305Granite-4.0-350M25.3%estimated ± 3.2 pp, medium confidence
306Granite-4.0-H-350M24.9%estimated ± 3.2 pp, medium confidence
307LFM2.5-VL-450M24.0%estimated ± 4.5 pp, medium confidence
308LFM2.5-230M23.7%estimated ± 4.5 pp, medium confidence
309MiniCPM5-1B20.2%estimated ± 12.1 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General