benchgap
Agentic · tools

GDPval-AA leaderboard

As of 2026-11-21, the highest measured score on GDPval-AA is 68.3% by Claude Opus 5.5. 146 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.270.6%estimated ± 1.7 pp, low confidence
2Claude Opus 5.568.3%measured
3Muse Wonder Full67.9%measured
4Kimi K3.1 Pro67.3%measured
5Claude Sonnet 5.566.9%measured
6Hy464.7%estimated ± 4.2 pp, high confidence
7Muse Wonder63.5%measured
8GLM-5.3-Flash63.0%estimated ± 5.6 pp, medium confidence
9Claude Fable 5.162.7%measured
10Gemini 4 Carbon62.7%estimated ± 1.7 pp, medium confidence
11GPT-6.2 Sol61.7%estimated ± 6.9 pp, low confidence
12Qwen4 Max61.7%estimated ± 8.6 pp, medium confidence
13Claude Opus 561.1%measured
14Inkling 2 Small61.0%estimated ± 6.2 pp, medium confidence
15Gemini 4 Neon60.9%estimated ± 6.4 pp, low confidence
16GPT-6.2 Astra60.9%estimated ± 6.4 pp, medium confidence
17Grok 4.760.7%measured
18MiMo-V2.6-Pro59.2%measured
19Gemini 4 Hydrogen59.2%estimated ± 1.7 pp, medium confidence
20Muse Spark 1.359.0%measured
21GPT-6 Cyber58.8%estimated ± 7.6 pp, low confidence
22Qwen3.8 Max Preview58.3%measured
23Claude Mythos 5.257.8%estimated ± 7.6 pp, low confidence
24LongCat-2.557.6%estimated ± 6.2 pp, medium confidence
25Muse Ember57.6%estimated ± 4.2 pp, high confidence
26GLM-5.357.5%measured
27ERNIE 5.5 Flash Preview57.4%estimated ± 4.6 pp, medium confidence
28GPT-5.6 Cyber57.0%estimated ± 7.6 pp, low confidence
29GPT-6.1 Terra56.5%estimated ± 1.7 pp, medium confidence
30Hy4 flash56.5%estimated ± 4.2 pp, high confidence
31Qwen3.8-Flash-Next56.4%measured
32Gemini 4 Argon56.2%measured
33Claude Haiku 5.556.0%measured
34Ultramarine 1 Flash56.0%estimated ± 6.2 pp, medium confidence
35Grok 4.655.9%measured
36Ling 3.1 Flash55.9%measured
37Claude Fable 555.5%measured
38GPT-5.6 Sol55.4%measured
39MiMo-V2.6-Flash55.4%measured
40Claude Mythos 5.155.3%estimated ± 7.6 pp, low confidence
41DeepSeek V4.1 Flash 101755.0%measured
42GPT-6.1 Sol54.8%measured
43DeepSeek V4 Pro54.5%measured
44Step 5 Preview54.3%measured
45GPT-6 Astra53.7%measured
46Fugu Cyber53.5%estimated ± 5.0 pp, high confidence
47Atria Dawn Preview53.2%estimated ± 5.0 pp, high confidence
48GPT-5.5 Pro53.1%estimated ± 7.8 pp, medium confidence
49Gemini 3.8 Flash Cyber53.0%estimated ± 5.0 pp, high confidence
50Qwen3.8 Max52.5%estimated ± 5.6 pp, medium confidence
51GPT-5.4 Pro51.8%estimated ± 7.8 pp, medium confidence
52Kimi K351.6%measured
53Claude Mythos 550.8%estimated ± 5.0 pp, high confidence
54Ornith-1.5-397B50.5%estimated ± 5.6 pp, medium confidence
55GPT-6 Sol50.4%measured
56Gemini 3.5 Flash Cyber50.3%estimated ± 5.0 pp, high confidence
57Claude Mythos Preview50.2%estimated ± 5.0 pp, high confidence
58Holo3-35B-A3B48.7%estimated ± 8.5 pp, medium confidence
59Muse Spark 1.248.6%measured
60Claude Sonnet 548.2%measured
61GPT-5.6 Luna48.0%measured
62MiniMax M3.1 Flash47.9%measured
63Claude Opus 4.847.7%measured
64GPT-5.6 Terra47.6%measured
65Inkling 2 Mini47.0%estimated ± 6.2 pp, medium confidence
66Gemini 3.8 Flash46.8%measured
67DeepSeek V4 Flash 073146.7%measured
68GPT-6 Luna46.6%measured
69Ling 3.0 Flash Fin46.3%estimated ± 6.7 pp, medium confidence
70Mistral Large 446.1%measured
71Ling 3.0 Flash Sante46.1%estimated ± 7.7 pp, medium confidence
72Qwen3.8-27B45.9%measured
73Claude Haiku 4.545.5%estimated ± 1.7 pp, low confidence
74Muse Spark Open45.1%measured
75Gemini 3.7 Flash44.6%measured
76Grok 4.544.2%measured
77GLM-5.243.6%measured
78Holo3-122B-A10B43.5%estimated ± 8.5 pp, medium confidence
79Sage-A1.543.0%estimated ± 7.7 pp, medium confidence
80Claude Opus 4.742.6%measured
81GPT-5.542.5%measured
82Gemini 3.5 Flash42.2%measured
83Hy4 preview41.0%estimated ± 4.6 pp, medium confidence
84Gemini 3.6 Flash39.2%measured
85BTL-338.6%estimated ± 5.5 pp, low confidence
86UI-Mate-27B38.0%estimated ± 8.5 pp, medium confidence
87Beam 1.0 501B-A23B FP837.9%estimated ± 7.7 pp, medium confidence
88Beam 1.0 501B-A23B NVFP437.9%estimated ± 7.7 pp, medium confidence
89GPT-5.437.3%measured
90MiniMax M337.1%measured
91GPT-5.236.8%estimated ± 6.2 pp, medium confidence
92GPT-5.3 Codex36.4%estimated ± 6.2 pp, medium confidence
93Hy3 Preview35.8%measured
94Muse Spark 1.135.6%measured
95Claude Opus 4.635.6%estimated ± 5.0 pp, high confidence
96Claude Sonnet 4.634.5%estimated ± 5.0 pp, high confidence
97Agents-A134.4%estimated ± 7.7 pp, medium confidence
98Apodex 1.134.4%measured
99Apodex 1.1 Mini34.4%measured
100Quasar 438B34.1%measured
101Beam 1.0 501B-A23B33.9%measured
102Ripple-65B Preview33.9%estimated ± 10.2 pp, low confidence
103GLM-5-Turbo33.8%estimated ± 11.0 pp, low confidence
104Nemotron 3.5 Super VL 120B A12B33.1%measured
105Nemotron 3.5 Super VL 120B A12B NVFP433.1%measured
106Nemotron 3 Ultra33.1%measured
107Ling 3.0 Flash VL33.0%measured
108MiMo-V2.532.8%estimated ± 8.0 pp, low confidence
109Claude Sonnet 4.532.3%estimated ± 6.2 pp, medium confidence
110Qwen3.7 Max31.4%measured
111dots3-note Preview31.4%estimated ± 5.6 pp, medium confidence
112GPT-5.1-Codex31.3%estimated ± 6.2 pp, medium confidence
113GPT-5.2-Codex31.2%estimated ± 6.2 pp, medium confidence
114Inkling-Small31.2%measured
115MiMo-V2.5-Pro31.0%measured
116GLM-5.130.8%measured
117Qwen 3.6 Max (preview)30.3%estimated ± 11.0 pp, low confidence
118Solar Pro 430.3%measured
119MiMo-V2-Pro30.1%estimated ± 8.0 pp, low confidence
120Solar Open 229.7%estimated ± 7.5 pp, low confidence
121Grok 4.329.2%measured
122Inkling28.6%measured
123Nemotron 3.5 Bolt 74B-A9B BF1628.5%estimated ± 7.8 pp, medium confidence
124Claude 4.1 Opus28.5%estimated ± 6.2 pp, medium confidence
125Hy328.2%measured
126Qwen3.5 397B27.7%estimated ± 7.8 pp, medium confidence
127Claude Opus 4.527.4%estimated ± 5.0 pp, high confidence
128Grok 4.127.1%estimated ± 8.0 pp, low confidence
129Grok 4.1 Fast (Reasoning)27.0%estimated ± 11.0 pp, low confidence
130Kimi K2.626.9%measured
131Kimi K2.7 Code26.9%measured
132Qwen3.5 Plus26.2%estimated ± 6.2 pp, medium confidence
133Agents-A1-4B26.2%estimated ± 7.8 pp, medium confidence
134Claude 4 Sonnet26.1%estimated ± 6.2 pp, medium confidence
135Laguna S 2.126.1%estimated ± 5.6 pp, low confidence
136GLM-525.9%estimated ± 5.0 pp, medium confidence
137Step 3.7 Flash25.8%measured
138GLM-4.725.7%measured
139GPT-5.4 mini25.6%measured
140Claude Opus 4.6 (Adaptive)25.6%estimated ± 11.0 pp, low confidence
141Qwen3.5-4B25.6%estimated ± 11.0 pp, low confidence
142MiniMax M2.725.5%measured
143Ornith-1.5-35B-A3B25.2%estimated ± 5.6 pp, low confidence
144Muse Spark24.9%measured
145Qwen3.6 Plus24.5%measured
146MiMo-V2-Omni24.5%estimated ± 11.0 pp, low confidence
147Qwen3.6-27B24.3%measured
148Gemini 3.5 Flash-Lite24.2%measured
149Mercury 324.1%estimated ± 12.2 pp, low confidence
150Nemotron 3.5 Bolt 74B-A9B24.0%measured
151Grok Build 0.123.4%estimated ± 8.8 pp, medium confidence
152A.X K223.0%measured
153Qwen3.5-27B22.9%estimated ± 7.8 pp, medium confidence
154Qwen3.5-35B-A3B22.9%estimated ± 7.8 pp, medium confidence
155Claude Opus 4.5 Thinking22.5%estimated ± 11.0 pp, low confidence
156GPT-5.4 nano22.4%measured
157Ling 3.0 Flash22.4%measured
158Grok 4.1 Fast21.9%estimated ± 8.8 pp, medium confidence
159BTL-421.4%estimated ± 5.5 pp, medium confidence
160Gemini 3 Flash21.4%estimated ± 6.2 pp, medium confidence
161Gemini 3 Pro21.4%estimated ± 6.2 pp, medium confidence
162GPT-520.9%measured
163UI-Mate-9B19.9%estimated ± 8.5 pp, medium confidence
164Qwen3.6-35B-A3B19.6%measured
165Ornith-1.5-9B19.6%estimated ± 5.6 pp, low confidence
166Spark-X2.6-4B19.5%estimated ± 5.5 pp, medium confidence
167Sarvam 105B19.4%estimated ± 7.8 pp, medium confidence
168LongCat-Flash-Lite-Sparse19.3%estimated ± 7.8 pp, medium confidence
169LFM3-1.7B19.0%estimated ± 5.5 pp, medium confidence
170Qwen3 Max19.0%estimated ± 8.8 pp, medium confidence
171Pokee-Isaac 28B18.5%estimated ± 5.5 pp, medium confidence
172Sarvam 30B18.2%estimated ± 7.8 pp, low confidence
173Grok 417.9%estimated ± 8.8 pp, medium confidence
174Kimi K2.517.0%measured
175GPT-5.116.3%measured
176GPT-5.1-Codex-Max15.9%estimated ± 11.0 pp, low confidence
177Qwen3.5-122B-A10B15.7%measured
178K-EXAONE 2.015.4%measured
179Gemini 3.1 Flash-Lite14.7%estimated ± 8.8 pp, medium confidence
180Grok 4.2014.6%estimated ± 8.8 pp, medium confidence
181Gemini 3.1 Pro14.6%measured
182Muse Glimmer 30B14.2%measured
183o313.8%estimated ± 11.0 pp, low confidence
184Ternary Bonsai 2 27B13.4%estimated ± 11.0 pp, low confidence
185Qwen3.7 Plus13.3%measured
186Mistral Medium 3.5 128B13.0%measured
187LFM2.5-24B-A2B11.7%estimated ± 5.5 pp, medium confidence
188GLM-4.610.9%estimated ± 11.0 pp, low confidence
189MiniCPM5-2B10.4%measured
190Grok Code Fast 110.1%estimated ± 11.0 pp, low confidence
191Claude 4 Opus Thinking8.6%estimated ± 11.0 pp, low confidence
192GLM-5V-Turbo8.5%estimated ± 8.8 pp, medium confidence
193DeepSeek V3.27.5%estimated ± 8.8 pp, low confidence
194Nemotron 3.5 Lightning 30B A3B NVFP46.9%measured
195LLaDA2.2-flash6.9%estimated ± 5.5 pp, medium confidence
196MiMo-V2-Flash6.4%measured
197Gemma 4 31B5.9%measured
198GPT-OSS 120B5.4%measured
199Grok 4 Fast (Reasoning)4.9%estimated ± 11.0 pp, low confidence
200GPT-4.14.3%estimated ± 8.8 pp, low confidence
201o13.7%estimated ± 11.0 pp, low confidence
202Granite 4.2 30B3.7%measured
203Kimi K23.3%estimated ± 11.0 pp, low confidence
204Gemma 4 26B A4B3.2%measured
205Ling 3.0 Tiny3.0%measured
206o4-mini2.0%estimated ± 11.0 pp, low confidence
207Nemotron 3.5 Bolt E4B BF161.9%estimated ± 11.0 pp, low confidence
208Claude 3.7 Sonnet Thinking1.8%estimated ± 11.0 pp, low confidence
209Nemotron 3.5 Bolt E4B1.4%estimated ± 11.0 pp, low confidence
210Claude 3.7 Sonnet1.1%estimated ± 11.0 pp, low confidence
211GLM-4.5-Air0.7%estimated ± 11.0 pp, low confidence
212Command A+0.7%measured
213DeepSeek-R10.2%estimated ± 11.0 pp, low confidence
214DeepSeek V3.10.1%estimated ± 11.0 pp, low confidence
215Solar Pro 20.1%estimated ± 11.0 pp, low confidence
216Mistral Large 20.1%estimated ± 11.0 pp, low confidence
217Gemini 2.0 Flash0.1%estimated ± 11.0 pp, low confidence
218o3-mini0.1%estimated ± 11.0 pp, low confidence
219Devstral Small0.0%estimated ± 11.0 pp, low confidence
220Magistral Small0.0%estimated ± 11.0 pp, low confidence
221Mistral Medium 30.0%estimated ± 11.0 pp, low confidence
222Magistral Medium0.0%estimated ± 11.0 pp, low confidence
223Llama 3.1 70B0.0%estimated ± 11.0 pp, low confidence
224Qwen3-Omni-30B-A3B-Thinking0.0%estimated ± 11.0 pp, low confidence
225Claude 3 Haiku0.0%estimated ± 11.0 pp, low confidence
226Exaone 4.0 1.2B0.0%estimated ± 11.0 pp, low confidence
227Granite-4.0-H-1B0.0%estimated ± 11.0 pp, low confidence
228Gemini 2.5 Flash-Lite0.0%estimated ± 11.0 pp, low confidence
229Llama 3.1 405B0.0%estimated ± 11.0 pp, low confidence
230Qwen3-Omni-30B-A3B-Instruct0.0%estimated ± 11.0 pp, low confidence
231Gemini 2.5 Flash0.0%estimated ± 11.0 pp, low confidence
232Granite-4.0-H-350M0.0%estimated ± 11.0 pp, low confidence
233Nova Pro0.0%estimated ± 11.0 pp, low confidence
234Granite-4.0-350M0.0%estimated ± 11.0 pp, low confidence
235Nemotron Ultra 253B0.0%estimated ± 11.0 pp, low confidence
236LFM2.5-VL-1.6B-Extract0.0%estimated ± 11.0 pp, low confidence
237Exaone 4.0 32B0.0%estimated ± 11.0 pp, low confidence
238Phi-40.0%estimated ± 11.0 pp, low confidence
239Phi-4-reasoning0.0%estimated ± 11.0 pp, low confidence
240Celeris-10.0%measured
241Claude 3.5 Haiku0.0%measured
242DeepSeek V30.0%measured
243DeepSeek V3 03240.0%measured
244Gemini 2.5 Pro0.0%measured
245Gemma 3 27B0.0%measured
246Gemma 4 12B0.0%measured
247Gemma 4 E2B0.0%measured
248Gemma 4 E4B0.0%measured
249GPT-4.1 mini0.0%measured
250GPT-4.1 nano0.0%measured
251GPT-4o0.0%measured
252GPT-4o mini0.0%measured
253GPT-OSS 20B0.0%measured
254Granite 4.2 3B0.0%measured
255Granite 4.2 8B0.0%measured
256K-Exaone0.0%measured
257LFM2.5-2.6B0.0%measured
258LFM2.5-8B-A1B0.0%estimated ± 5.5 pp, low confidence
259LFM2.5-VL-3B0.0%estimated ± 5.5 pp, low confidence
260Ling 2.6 Flash0.0%measured
261LLaDA2.2-mini0.0%estimated ± 5.5 pp, low confidence
262Llama 3.3 70B0.0%measured
263Llama 4 Maverick0.0%measured
264Llama 4 Scout0.0%measured
265Mellum2-12B-A2.5B-Instruct0.0%estimated ± 5.5 pp, low confidence
266Mellum2-12B-A2.5B-Thinking0.0%estimated ± 5.5 pp, low confidence
267Mercury 2.50.0%measured
268Mistral Large 30.0%measured
269Mistral Small 3.10.0%measured
270Mistral Small 40.0%measured
271Nemotron 3 Nano 30B0.0%measured
272Nemotron 3 Nano Omni 30B A3B0.0%measured
273North Mini Code0.0%measured
274Solar Pro 30.0%measured
275Trinity-Large-Preview0.0%measured
276Trinity-Large-Thinking0.0%measured
277Ultravox v0.6 Llama 3.3 70B0.0%measured
278ZAYA1-8B0.0%estimated ± 5.5 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General