benchgap
Agentic · tools

BrowseComp leaderboard

As of 2026-11-21, the highest measured score on BrowseComp is 92.5% by Atria Dawn Preview. 209 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Muse Wonder Full97.9%estimated ± 6.2 pp, low confidence
2Gemini 4 Carbon96.6%estimated ± 3.0 pp, low confidence
3Muse Wonder95.4%estimated ± 6.2 pp, medium confidence
4GPT-6 Cyber95.2%estimated ± 3.0 pp, low confidence
5Qwen4 Max93.0%estimated ± 4.2 pp, medium confidence
6GPT-6.1 Terra93.0%estimated ± 5.0 pp, low confidence
7GPT-6.2 Astra92.7%estimated ± 2.9 pp, low confidence
8Atria Dawn Preview92.5%measured
9Claude Mythos 5.292.3%estimated ± 3.0 pp, medium confidence
10GPT-5.6 Sol92.2%measured
11Kimi K3.1 Pro92.1%measured
12LongCat-2.591.9%measured
13Ling 3.1 Flash91.7%measured
14Gemini 4 Argon91.7%estimated ± 1.9 pp, high confidence
15Muse Spark 1.391.7%estimated ± 1.9 pp, high confidence
16MiniMax M3.1 Flash91.6%estimated ± 1.9 pp, high confidence
17GPT-6 Sol91.6%estimated ± 1.9 pp, high confidence
18Gemini 3.8 Flash91.6%estimated ± 1.9 pp, high confidence
19GPT-6 Astra91.5%measured
20GPT-6.1 Sol91.4%estimated ± 3.0 pp, medium confidence
21GPT-6.2 Sol91.4%estimated ± 2.9 pp, low confidence
22Ultramarine 1 Flash91.3%estimated ± 6.2 pp, medium confidence
23Kimi K391.2%measured
24Claude Fable 591.2%estimated ± 2.9 pp, high confidence
25Claude Fable 5.291.0%estimated ± 2.7 pp, medium confidence
26Gemini 4 Neon90.9%estimated ± 11.3 pp, low confidence
27Gemini 4 Hydrogen90.9%estimated ± 11.3 pp, low confidence
28Hy4 preview90.8%estimated ± 1.9 pp, medium confidence
29Claude Opus 590.8%measured
30Claude Sonnet 5.590.6%estimated ± 2.7 pp, medium confidence
31GPT-5.6 Cyber90.6%estimated ± 3.0 pp, medium confidence
32Grok 4.790.6%estimated ± 4.0 pp, high confidence
33Gemini 3.5 Flash90.2%estimated ± 4.2 pp, medium confidence
34Inkling 2 Small90.2%measured
35GPT-5.5 Pro90.1%measured
36Hy489.9%estimated ± 4.8 pp, low confidence
37Hy4 flash89.9%estimated ± 4.8 pp, medium confidence
38GPT-5.4 Pro89.3%measured
39GLM-5.3-Flash89.0%estimated ± 4.0 pp, high confidence
40LFM3-1.7B88.9%estimated ± 6.7 pp, low confidence
41Qwen3.8 Max Preview88.8%estimated ± 4.2 pp, medium confidence
42Step 5 Preview88.7%measured
43Fugu Cyber88.4%estimated ± 5.2 pp, medium confidence
44Claude Mythos 5.188.2%estimated ± 3.0 pp, medium confidence
45Claude Mythos 588.0%measured
46Gemini 3.8 Flash Cyber88.0%estimated ± 5.2 pp, medium confidence
47Ripple-65B Preview87.6%estimated ± 6.7 pp, low confidence
48GPT-5.6 Terra87.5%measured
49ERNIE 5.5 Flash Preview87.4%measured
50Grok 4.587.2%estimated ± 4.7 pp, high confidence
51Ornith-1.5-397B86.6%measured
52MiMo-V2.6-Pro86.6%estimated ± 3.0 pp, medium confidence
53Claude Mythos Preview86.5%estimated ± 3.0 pp, medium confidence
54Gemini 3.5 Flash Cyber86.4%estimated ± 5.2 pp, medium confidence
55Muse Spark Open86.2%estimated ± 6.0 pp, medium confidence
56DeepSeek V4.1 Flash 101785.9%estimated ± 3.0 pp, medium confidence
57GLM-5.385.8%estimated ± 3.0 pp, medium confidence
58Claude Opus 5.585.3%estimated ± 1.9 pp, high confidence
59GLM-5.285.3%estimated ± 4.2 pp, high confidence
60Claude Opus 4.6 (Adaptive)85.0%estimated ± 4.2 pp, high confidence
61Muse Spark 1.284.9%estimated ± 9.0 pp, medium confidence
62GPT-6 Luna84.8%estimated ± 3.0 pp, medium confidence
63Qwen3.8-27B84.8%estimated ± 4.0 pp, high confidence
64Claude Sonnet 584.7%measured
65Mistral Large 484.5%measured
66LFM2.5-24B-A2B84.5%estimated ± 6.7 pp, low confidence
67Gemini 3.7 Flash84.5%estimated ± 1.9 pp, high confidence
68Gemini 3.1 Pro84.5%estimated ± 4.2 pp, high confidence
69GPT-5.584.4%measured
70Claude Opus 4.884.3%measured
71Grok 4.684.1%estimated ± 2.9 pp, high confidence
72Apodex 1.184.0%estimated ± 4.2 pp, high confidence
73Apodex 1.1 Mini84.0%estimated ± 4.2 pp, high confidence
74Gemini 3.6 Flash83.8%estimated ± 4.7 pp, high confidence
75Holo3-35B-A3B83.8%estimated ± 4.7 pp, high confidence
76Holo3-122B-A10B83.8%estimated ± 4.7 pp, high confidence
77UI-Mate-27B83.8%estimated ± 4.7 pp, high confidence
78Gemini 3.5 Flash-Lite83.8%estimated ± 4.7 pp, high confidence
79Claude Opus 4.683.7%measured
80MiniMax M383.5%measured
81DeepSeek V4 Pro83.4%measured
82dots3-note Preview83.3%measured
83GPT-5.6 Luna83.3%measured
84MiMo-V2.6-Flash83.3%estimated ± 3.0 pp, medium confidence
85Kimi K2.683.2%measured
86Claude Fable 5.183.0%estimated ± 1.9 pp, high confidence
87Qwen3.8-Flash-Next82.9%estimated ± 1.9 pp, high confidence
88Qwen3.8 Max82.9%estimated ± 1.9 pp, high confidence
89Muse Spark 1.182.9%estimated ± 1.9 pp, high confidence
90Claude Sonnet 4.682.9%estimated ± 1.9 pp, high confidence
91Qwen3.7 Plus82.9%estimated ± 1.9 pp, medium confidence
92GPT-5.482.7%measured
93GPT-5.4 mini82.3%estimated ± 4.2 pp, high confidence
94Claude Haiku 5.582.2%estimated ± 7.1 pp, medium confidence
95Ling 3.0 Flash Fin81.7%estimated ± 4.2 pp, high confidence
96Sage-A1.581.7%measured
97Qwen3.7 Max81.4%estimated ± 5.6 pp, medium confidence
98MiMo-V2.581.3%estimated ± 5.6 pp, medium confidence
99Ling 3.0 Flash Sante80.2%measured
100GPT-5.4 nano79.9%estimated ± 4.2 pp, high confidence
101Quasar 438B79.8%estimated ± 9.0 pp, medium confidence
102MiMo-V2-Pro79.6%estimated ± 5.6 pp, medium confidence
103Claude Opus 4.779.3%measured
104Inkling 2 Mini79.0%measured
105Kimi K2.7 Code78.6%estimated ± 7.8 pp, medium confidence
106Nemotron 3.5 Super VL 120B A12B78.0%measured
107GPT-5.3-Codex-Spark77.6%estimated ± 11.3 pp, low confidence
108Muse Ember77.6%estimated ± 11.3 pp, low confidence
109Nemotron 3.5 Super VL 120B A12B NVFP477.5%measured
110Inkling-Small77.4%measured
111Inkling77.1%measured
112Beam 1.0 501B-A23B76.8%measured
113Beam 1.0 501B-A23B FP876.7%measured
114Mercury 376.4%estimated ± 6.8 pp, low confidence
115Step 3.7 Flash75.8%measured
116Beam 1.0 501B-A23B NVFP475.6%measured
117Agents-A175.5%measured
118Ling 3.0 Flash VL75.5%estimated ± 9.2 pp, medium confidence
119Hy375.1%estimated ± 9.0 pp, medium confidence
120GLM-5-Turbo74.1%estimated ± 13.0 pp, low confidence
121Claude Opus 4.573.4%estimated ± 4.3 pp, high confidence
122DeepSeek V4 Flash 073173.2%measured
123Qwen 3.6 Max (preview)73.2%estimated ± 13.0 pp, low confidence
124Qwen3.6-27B72.4%estimated ± 6.6 pp, medium confidence
125Grok 4.1 Fast (Reasoning)72.2%estimated ± 13.0 pp, low confidence
126Ling 3.0 Flash72.2%measured
127GPT-5.1-Codex71.9%estimated ± 6.2 pp, medium confidence
128Grok 4.371.8%estimated ± 4.2 pp, high confidence
129GPT-5.2-Codex71.8%estimated ± 6.2 pp, medium confidence
130Qwen3.5-4B71.8%estimated ± 13.0 pp, low confidence
131Laguna S 2.171.7%estimated ± 6.0 pp, medium confidence
132MiMo-V2-Omni71.4%estimated ± 13.0 pp, low confidence
133Muse Glimmer 30B71.2%estimated ± 1.9 pp, low confidence
134Claude Opus 4.5 Thinking70.8%estimated ± 13.0 pp, low confidence
135LFM2.5-8B-A1B70.2%estimated ± 13.0 pp, low confidence
136Nemotron 3.5 Bolt 74B-A9B BF1670.0%measured
137Qwen3.6 Plus69.8%estimated ± 4.3 pp, high confidence
138Claude 4.1 Opus69.6%estimated ± 6.2 pp, medium confidence
139Qwen3.5 397B69.0%measured
140Grok Build 0.168.2%estimated ± 6.6 pp, medium confidence
141GPT-5.1-Codex-Max68.1%estimated ± 13.0 pp, low confidence
142GLM-5.168.0%measured
143Qwen3.5 Plus67.8%estimated ± 6.2 pp, medium confidence
144Claude 4 Sonnet67.7%estimated ± 6.2 pp, medium confidence
145Ornith-1.5-35B-A3B67.6%measured
146Gemma 4 31B67.4%estimated ± 5.7 pp, medium confidence
147o367.1%estimated ± 13.0 pp, low confidence
148Grok 4.1 Fast66.9%estimated ± 6.6 pp, medium confidence
149Ternary Bonsai 2 27B66.9%estimated ± 13.0 pp, low confidence
150Agents-A1-4B66.8%measured
151Nemotron 3.5 Bolt 74B-A9B66.8%measured
152Claude Haiku 4.566.5%estimated ± 6.2 pp, medium confidence
153A.X K265.9%estimated ± 9.2 pp, medium confidence
154GPT-5.265.8%measured
155GLM-4.665.4%estimated ± 13.0 pp, low confidence
156Grok Code Fast 164.9%estimated ± 13.0 pp, low confidence
157Qwen3 Max64.3%estimated ± 6.6 pp, medium confidence
158UI-Mate-9B64.1%estimated ± 4.7 pp, high confidence
159Gemini 3 Flash64.0%estimated ± 6.2 pp, medium confidence
160Gemini 3 Pro64.0%estimated ± 6.2 pp, medium confidence
161Claude 4 Opus Thinking63.8%estimated ± 13.0 pp, low confidence
162Qwen3.5-122B-A10B63.8%measured
163GPT-5.3 Codex63.6%estimated ± 4.7 pp, high confidence
164Claude Sonnet 4.563.5%estimated ± 4.7 pp, high confidence
165Grok 463.3%estimated ± 6.6 pp, medium confidence
166Gemini 2.5 Pro63.1%estimated ± 6.6 pp, medium confidence
167GPT-5.162.5%estimated ± 6.6 pp, medium confidence
168GPT-562.5%estimated ± 6.2 pp, low confidence
169GLM-562.0%measured
170LLaDA2.2-flash61.3%estimated ± 7.8 pp, medium confidence
171Qwen3.5-27B61.0%measured
172Qwen3.5-35B-A3B61.0%measured
173Mistral Medium 3.5 128B61.0%estimated ± 6.6 pp, medium confidence
174Kimi K2.560.6%measured
175Gemini 3.1 Flash-Lite60.5%estimated ± 6.6 pp, medium confidence
176Grok 4.2060.4%estimated ± 6.6 pp, medium confidence
177MiniMax M2.760.2%estimated ± 4.2 pp, medium confidence
178Muse Spark60.1%estimated ± 5.2 pp, medium confidence
179Grok 4 Fast (Reasoning)60.0%estimated ± 13.0 pp, low confidence
180Hy3 Preview59.4%estimated ± 6.6 pp, medium confidence
181o158.3%estimated ± 13.0 pp, low confidence
182Kimi K257.5%estimated ± 13.0 pp, low confidence
183Qwen3.6-35B-A3B57.4%estimated ± 4.3 pp, medium confidence
184Ornith-1.5-9B56.4%measured
185Trinity-Large-Thinking56.2%estimated ± 6.6 pp, medium confidence
186LLaDA2.2-mini55.5%estimated ± 13.0 pp, low confidence
187GLM-5V-Turbo55.0%estimated ± 6.6 pp, medium confidence
188K-EXAONE 2.054.7%estimated ± 9.2 pp, medium confidence
189o4-mini54.3%estimated ± 13.0 pp, low confidence
190DeepSeek V3.254.1%estimated ± 6.6 pp, medium confidence
191Nemotron 3.5 Bolt E4B BF1654.0%estimated ± 13.0 pp, low confidence
192Claude 3.7 Sonnet Thinking53.8%estimated ± 13.0 pp, low confidence
193Spark-X2.6-4B53.1%measured
194Nemotron 3.5 Bolt E4B52.3%estimated ± 13.0 pp, low confidence
195GLM-4.752.0%measured
196GPT-4.151.3%estimated ± 6.6 pp, low confidence
197Claude 3.7 Sonnet50.9%estimated ± 13.0 pp, low confidence
198Sarvam 105B49.5%measured
199Solar Pro 449.2%measured
200Solar Open 249.2%estimated ± 4.8 pp, low confidence
201LongCat-Flash-Lite-Sparse48.6%measured
202GLM-4.5-Air48.5%estimated ± 13.0 pp, low confidence
203Nemotron 3 Ultra44.4%measured
204Grok 4.144.3%estimated ± 5.6 pp, low confidence
205MiniCPM5-2B43.8%estimated ± 9.2 pp, medium confidence
206DeepSeek-R141.2%estimated ± 13.0 pp, low confidence
207DeepSeek V3.139.8%estimated ± 13.0 pp, low confidence
208Solar Pro 237.4%estimated ± 13.0 pp, low confidence
209Nemotron 3.5 Lightning 30B A3B NVFP436.8%measured
210Mistral Large 236.3%estimated ± 13.0 pp, low confidence
211Sarvam 30B35.5%measured
212Gemini 2.0 Flash35.3%estimated ± 13.0 pp, low confidence
213o3-mini34.5%estimated ± 13.0 pp, low confidence
214Devstral Small34.3%estimated ± 13.0 pp, low confidence
215Magistral Small32.6%estimated ± 13.0 pp, low confidence
216MiMo-V2-Flash31.7%estimated ± 9.2 pp, low confidence
217Mistral Medium 330.4%estimated ± 13.0 pp, low confidence
218GPT-OSS 120B29.5%estimated ± 4.2 pp, medium confidence
219Magistral Medium29.2%estimated ± 13.0 pp, low confidence
220Granite 4.2 8B28.5%estimated ± 9.0 pp, low confidence
221Command A+28.0%estimated ± 9.0 pp, low confidence
222Llama 3.1 70B27.7%estimated ± 13.0 pp, low confidence
223Qwen3-Omni-30B-A3B-Thinking27.4%estimated ± 13.0 pp, low confidence
224Claude 3 Haiku27.2%estimated ± 13.0 pp, low confidence
225Exaone 4.0 1.2B26.5%estimated ± 13.0 pp, low confidence
226Granite-4.0-H-1B25.6%estimated ± 13.0 pp, low confidence
227Gemini 2.5 Flash-Lite24.9%estimated ± 13.0 pp, low confidence
228Llama 3.1 405B24.9%estimated ± 13.0 pp, low confidence
229MiMo-V2.5-Pro24.4%estimated ± 4.2 pp, medium confidence
230Qwen3-Omni-30B-A3B-Instruct22.1%estimated ± 13.0 pp, low confidence
231Granite 4.2 30B20.8%estimated ± 9.2 pp, low confidence
232Mistral Large 320.7%estimated ± 9.0 pp, low confidence
233Gemini 2.5 Flash20.3%estimated ± 13.0 pp, low confidence
234Granite-4.0-H-350M20.0%estimated ± 13.0 pp, low confidence
235Nova Pro19.3%estimated ± 13.0 pp, low confidence
236Gemma 4 26B A4B18.4%estimated ± 9.2 pp, low confidence
237Granite-4.0-350M18.3%estimated ± 13.0 pp, low confidence
238Ling 3.0 Tiny17.4%estimated ± 9.2 pp, low confidence
239Nemotron Ultra 253B16.1%estimated ± 13.0 pp, low confidence
240Mistral Small 413.3%estimated ± 9.0 pp, low confidence
241LFM2.5-VL-1.6B-Extract12.4%estimated ± 13.0 pp, low confidence
242Trinity-Large-Preview11.0%estimated ± 9.0 pp, low confidence
243Nemotron 3 Nano 30B9.6%estimated ± 9.0 pp, low confidence
244GPT-OSS 20B8.5%estimated ± 4.2 pp, medium confidence
245DeepSeek V37.7%estimated ± 9.0 pp, low confidence
246Celeris-16.4%estimated ± 9.0 pp, low confidence
247Exaone 4.0 32B6.3%estimated ± 13.0 pp, low confidence
248Llama 4 Maverick6.2%estimated ± 9.0 pp, low confidence
249Llama 4 Scout5.5%estimated ± 9.0 pp, low confidence
250Gemma 3 27B1.5%estimated ± 9.0 pp, low confidence
251Phi-40.0%estimated ± 13.0 pp, low confidence
252Phi-4-reasoning0.0%estimated ± 13.0 pp, low confidence
253Claude 3.5 Haiku0.0%estimated ± 9.2 pp, low confidence
254DeepSeek V3 03240.0%estimated ± 9.2 pp, low confidence
255Gemma 4 12B0.0%estimated ± 9.2 pp, low confidence
256Gemma 4 E2B0.0%estimated ± 9.2 pp, low confidence
257Gemma 4 E4B0.0%estimated ± 9.2 pp, low confidence
258GPT-4.1 mini0.0%estimated ± 9.2 pp, low confidence
259GPT-4.1 nano0.0%estimated ± 9.2 pp, low confidence
260GPT-4o0.0%estimated ± 9.2 pp, low confidence
261GPT-4o mini0.0%estimated ± 9.2 pp, low confidence
262Granite 4.2 3B0.0%estimated ± 9.2 pp, low confidence
263K-Exaone0.0%estimated ± 9.2 pp, low confidence
264LFM2.5-2.6B0.0%estimated ± 9.2 pp, low confidence
265Ling 2.6 Flash0.0%estimated ± 9.2 pp, low confidence
266Llama 3.3 70B0.0%estimated ± 9.2 pp, low confidence
267Mercury 2.50.0%estimated ± 9.2 pp, low confidence
268Mistral Small 3.10.0%estimated ± 9.2 pp, low confidence
269Nemotron 3 Nano Omni 30B A3B0.0%estimated ± 9.2 pp, low confidence
270North Mini Code0.0%estimated ± 9.2 pp, low confidence
271Solar Pro 30.0%estimated ± 9.2 pp, low confidence
272Ultravox v0.6 Llama 3.3 70B0.0%estimated ± 9.2 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General