benchgap
Agentic · tools

DeepSearchQA leaderboard

As of 2026-11-21, the highest measured score on DeepSearchQA is 96.0% by Atria Dawn Preview. 169 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Qwen4 Max100.0%estimated ± 7.9 pp, low confidence
2Fugu Cyber100.0%estimated ± 11.5 pp, low confidence
3Atria Dawn Preview96.0%measured
4Claude Opus 595.0%measured
5Kimi K395.0%measured
6Claude Fable 5.194.3%estimated ± 4.5 pp, medium confidence
7Muse Wonder Full93.9%estimated ± 8.2 pp, low confidence
8GLM-5.393.8%estimated ± 4.5 pp, high confidence
9Grok 4.693.8%estimated ± 4.5 pp, high confidence
10Kimi K3.1 Pro93.7%estimated ± 6.8 pp, low confidence
11GPT-6 Astra93.6%estimated ± 4.5 pp, high confidence
12LongCat-2.593.5%estimated ± 6.8 pp, low confidence
13Claude Fable 593.5%estimated ± 4.5 pp, high confidence
14GPT-5.6 Sol93.4%estimated ± 4.5 pp, high confidence
15Ling 3.1 Flash93.3%estimated ± 6.8 pp, low confidence
16Qwen3.8 Max Preview93.3%estimated ± 4.5 pp, high confidence
17DeepSeek V4 Pro93.3%estimated ± 4.5 pp, high confidence
18Claude Opus 4.893.1%measured
19Qwen3.8-27B92.9%estimated ± 4.5 pp, high confidence
20Step 3.7 Flash92.8%measured
21Claude Sonnet 592.5%estimated ± 4.5 pp, high confidence
22Kimi K2.692.5%measured
23Muse Spark 1.292.5%estimated ± 4.5 pp, high confidence
24GPT-5.6 Terra92.4%estimated ± 4.5 pp, high confidence
25Apodex 1.192.4%measured
26Muse Wonder92.3%estimated ± 8.2 pp, low confidence
27GPT-5.6 Luna92.2%estimated ± 4.5 pp, high confidence
28GPT-5.5 Pro92.2%estimated ± 6.8 pp, low confidence
29Grok 4.592.1%estimated ± 4.5 pp, high confidence
30dots3-note Preview92.1%measured
31DeepSeek V4 Flash 073192.1%estimated ± 4.5 pp, high confidence
32Gemini 3.8 Flash91.9%estimated ± 4.5 pp, high confidence
33GPT-5.4 Pro91.6%estimated ± 6.8 pp, low confidence
34Claude Opus 4.791.6%estimated ± 4.5 pp, high confidence
35GLM-5.291.6%estimated ± 4.5 pp, high confidence
36Muse Spark Open91.3%estimated ± 7.3 pp, low confidence
37Qwen3.8 Max91.3%estimated ± 7.9 pp, low confidence
38Step 5 Preview91.2%estimated ± 6.8 pp, low confidence
39GPT-5.591.1%estimated ± 4.5 pp, high confidence
40Hy4 preview91.0%estimated ± 7.3 pp, low confidence
41Gemini 3.7 Flash90.9%estimated ± 4.5 pp, high confidence
42Inkling 2 Small90.9%measured
43Ling 3.0 Flash Sante90.8%measured
44Claude Mythos 590.7%estimated ± 6.8 pp, low confidence
45ERNIE 5.5 Flash Preview90.3%estimated ± 6.8 pp, low confidence
46Gemini 3.8 Flash Cyber90.0%estimated ± 11.5 pp, low confidence
47Ultramarine 1 Flash90.0%estimated ± 8.2 pp, low confidence
48Quasar 438B89.9%estimated ± 4.5 pp, high confidence
49Ornith-1.5-397B89.7%estimated ± 6.8 pp, low confidence
50Claude Opus 5.589.7%estimated ± 9.6 pp, low confidence
51Claude Sonnet 5.589.7%estimated ± 9.6 pp, low confidence
52Grok 4.789.6%estimated ± 9.6 pp, medium confidence
53Gemini 4 Argon89.6%estimated ± 9.6 pp, medium confidence
54Claude Haiku 5.589.6%estimated ± 9.6 pp, medium confidence
55DeepSeek V4.1 Flash 101789.6%estimated ± 9.6 pp, medium confidence
56GPT-6.1 Sol89.6%estimated ± 9.6 pp, medium confidence
57GPT-6 Sol89.5%estimated ± 9.6 pp, medium confidence
58Muse Spark 1.389.4%measured
59MiniMax M3.1 Flash89.4%estimated ± 9.6 pp, medium confidence
60MiniMax M389.4%estimated ± 4.5 pp, high confidence
61GPT-6 Luna89.3%estimated ± 9.6 pp, medium confidence
62Gemini 3.6 Flash89.2%estimated ± 4.5 pp, high confidence
63Mistral Large 488.2%estimated ± 6.8 pp, low confidence
64Holo3-35B-A3B88.2%estimated ± 7.9 pp, low confidence
65Gemini 3.5 Flash88.1%estimated ± 4.5 pp, high confidence
66Apodex 1.1 Mini88.0%estimated ± 9.6 pp, medium confidence
67MiMo-V2.6-Pro87.7%estimated ± 7.9 pp, low confidence
68Qwen3.8-Flash-Next87.7%estimated ± 8.2 pp, low confidence
69Ling 3.0 Flash VL87.7%estimated ± 9.6 pp, medium confidence
70Hy387.4%estimated ± 4.5 pp, high confidence
71Hy3 Preview87.4%estimated ± 4.5 pp, high confidence
72GLM-5.187.2%estimated ± 4.5 pp, high confidence
73Inkling-Small87.1%estimated ± 4.5 pp, high confidence
74Inkling86.8%estimated ± 4.5 pp, high confidence
75MiMo-V2.6-Flash86.7%estimated ± 7.9 pp, low confidence
76Qwen3.7 Max86.6%estimated ± 4.5 pp, high confidence
77Sage-A1.586.4%estimated ± 6.8 pp, low confidence
78MiMo-V2.5-Pro85.9%estimated ± 4.5 pp, high confidence
79Kimi K2.7 Code85.8%estimated ± 4.5 pp, high confidence
80Nemotron 3 Ultra85.4%estimated ± 4.5 pp, high confidence
81Inkling 2 Mini85.3%measured
82Holo3-122B-A10B85.1%estimated ± 7.9 pp, low confidence
83Ling 3.0 Flash84.9%estimated ± 4.5 pp, high confidence
84Muse Spark 1.184.9%measured
85Qwen3.6-27B84.3%estimated ± 4.5 pp, high confidence
86Nemotron 3.5 Super VL 120B A12B84.1%estimated ± 6.8 pp, low confidence
87Qwen3.7 Plus84.1%estimated ± 4.5 pp, high confidence
88GPT-5.4 mini84.0%estimated ± 4.5 pp, high confidence
89Nemotron 3.5 Super VL 120B A12B NVFP483.8%estimated ± 6.8 pp, low confidence
90UI-Mate-27B83.6%estimated ± 7.9 pp, low confidence
91A.X K282.8%estimated ± 9.6 pp, medium confidence
92Agents-A182.7%estimated ± 6.8 pp, low confidence
93GPT-5.4 nano82.6%estimated ± 4.5 pp, high confidence
94Grok 4.382.2%estimated ± 4.5 pp, high confidence
95MiniMax M2.781.8%estimated ± 4.5 pp, high confidence
96Ripple-65B Preview81.1%estimated ± 7.3 pp, low confidence
97Claude Opus 4.581.1%estimated ± 7.3 pp, low confidence
98LFM3-1.7B81.1%estimated ± 7.3 pp, low confidence
99LLaDA2.2-flash81.1%estimated ± 7.3 pp, low confidence
100Qwen3.6 Plus81.1%estimated ± 7.3 pp, low confidence
101Solar Open 281.1%estimated ± 7.3 pp, low confidence
102Beam 1.0 501B-A23B NVFP481.1%measured
103Gemini 3.5 Flash-Lite80.9%estimated ± 4.5 pp, high confidence
104Beam 1.0 501B-A23B80.5%measured
105MiMo-V2.580.5%estimated ± 10.2 pp, low confidence
106Nemotron 3.5 Bolt 74B-A9B BF1680.0%estimated ± 6.8 pp, low confidence
107Qwen3.6-35B-A3B80.0%estimated ± 4.5 pp, high confidence
108Beam 1.0 501B-A23B FP879.6%measured
109Qwen3.5 397B79.6%estimated ± 6.8 pp, low confidence
110Ornith-1.5-35B-A3B79.1%estimated ± 6.8 pp, low confidence
111Agents-A1-4B78.8%estimated ± 6.8 pp, low confidence
112Nemotron 3.5 Bolt 74B-A9B78.7%estimated ± 6.8 pp, low confidence
113GPT-5.278.4%estimated ± 6.8 pp, low confidence
114GPT-5.1-Codex78.4%estimated ± 8.2 pp, low confidence
115GPT-5.2-Codex78.3%estimated ± 8.2 pp, low confidence
116GLM-577.3%estimated ± 6.8 pp, low confidence
117Kimi K2.577.1%measured
118Claude 4.1 Opus77.1%estimated ± 8.2 pp, low confidence
119Qwen3.5-27B77.0%estimated ± 6.8 pp, low confidence
120Qwen3.5-35B-A3B77.0%estimated ± 6.8 pp, low confidence
121MiMo-V2-Pro77.0%estimated ± 10.2 pp, low confidence
122Gemini 3.5 Flash Cyber76.8%estimated ± 11.5 pp, low confidence
123Claude Mythos Preview76.8%estimated ± 11.5 pp, low confidence
124UI-Mate-9B76.2%estimated ± 7.9 pp, low confidence
125Ornith-1.5-9B76.1%estimated ± 6.8 pp, low confidence
126Qwen3.5 Plus76.0%estimated ± 8.2 pp, low confidence
127Claude 4 Sonnet76.0%estimated ± 8.2 pp, low confidence
128Spark-X2.6-4B75.6%estimated ± 6.8 pp, low confidence
129GLM-4.775.4%estimated ± 6.8 pp, low confidence
130GPT-5.3 Codex75.4%estimated ± 7.9 pp, low confidence
131Claude Haiku 4.575.2%estimated ± 8.2 pp, low confidence
132Sarvam 105B75.2%estimated ± 6.8 pp, low confidence
133Solar Pro 475.1%estimated ± 6.8 pp, low confidence
134LongCat-Flash-Lite-Sparse75.1%estimated ± 6.8 pp, low confidence
135GPT-5.175.0%estimated ± 9.6 pp, medium confidence
136Muse Spark74.8%measured
137Muse Glimmer 30B74.6%measured
138Sarvam 30B74.5%estimated ± 6.8 pp, low confidence
139Claude Sonnet 4.573.9%estimated ± 7.9 pp, low confidence
140Gemini 3 Flash73.8%estimated ± 8.2 pp, low confidence
141Gemini 3 Pro73.8%estimated ± 8.2 pp, low confidence
142Claude Opus 4.673.7%measured
143GPT-5.473.6%measured
144K-EXAONE 2.073.6%estimated ± 9.6 pp, medium confidence
145Claude Sonnet 4.673.4%estimated ± 7.5 pp, low confidence
146GPT-572.9%estimated ± 8.2 pp, low confidence
147Grok 4.172.6%estimated ± 10.2 pp, low confidence
148Qwen3.5-122B-A10B71.8%estimated ± 4.5 pp, medium confidence
149Mercury 371.3%measured
150Mistral Medium 3.5 128B71.1%estimated ± 4.5 pp, medium confidence
151Gemini 3.1 Pro69.7%measured
152Gemma 4 31B63.9%estimated ± 4.5 pp, medium confidence
153MiniCPM5-2B63.3%estimated ± 9.6 pp, medium confidence
154Grok 4.2062.8%measured
155GPT-OSS 120B62.0%estimated ± 4.5 pp, medium confidence
156Nemotron 3.5 Lightning 30B A3B NVFP461.7%estimated ± 4.5 pp, medium confidence
157MiMo-V2-Flash52.9%estimated ± 9.6 pp, medium confidence
158Granite 4.2 8B49.2%estimated ± 4.5 pp, medium confidence
159Command A+48.7%estimated ± 4.5 pp, medium confidence
160Gemini 2.5 Pro48.1%estimated ± 4.5 pp, medium confidence
161Granite 4.2 30B45.2%estimated ± 9.6 pp, medium confidence
162Gemma 4 26B A4B43.8%estimated ± 9.6 pp, medium confidence
163Ling 3.0 Tiny43.2%estimated ± 9.6 pp, medium confidence
164Mistral Large 338.9%estimated ± 4.5 pp, medium confidence
165Claude 3.5 Haiku34.7%estimated ± 9.6 pp, medium confidence
166DeepSeek V3 032434.7%estimated ± 9.6 pp, medium confidence
167Gemma 4 12B34.7%estimated ± 9.6 pp, medium confidence
168Gemma 4 E2B34.7%estimated ± 9.6 pp, medium confidence
169Gemma 4 E4B34.7%estimated ± 9.6 pp, medium confidence
170GPT-4.1 mini34.7%estimated ± 9.6 pp, medium confidence
171GPT-4.1 nano34.7%estimated ± 9.6 pp, medium confidence
172GPT-4o34.7%estimated ± 9.6 pp, medium confidence
173GPT-4o mini34.7%estimated ± 9.6 pp, medium confidence
174Granite 4.2 3B34.7%estimated ± 9.6 pp, medium confidence
175K-Exaone34.7%estimated ± 9.6 pp, medium confidence
176LFM2.5-2.6B34.7%estimated ± 9.6 pp, medium confidence
177Ling 2.6 Flash34.7%estimated ± 9.6 pp, medium confidence
178Llama 3.3 70B34.7%estimated ± 9.6 pp, medium confidence
179Mistral Small 3.134.7%estimated ± 9.6 pp, medium confidence
180Nemotron 3 Nano Omni 30B A3B34.7%estimated ± 9.6 pp, medium confidence
181North Mini Code34.7%estimated ± 9.6 pp, medium confidence
182Solar Pro 334.7%estimated ± 9.6 pp, medium confidence
183Ultravox v0.6 Llama 3.3 70B34.7%estimated ± 9.6 pp, medium confidence
184Mercury 2.534.0%measured
185Mistral Small 427.3%estimated ± 4.5 pp, medium confidence
186GPT-OSS 20B26.7%estimated ± 4.5 pp, medium confidence
187Trinity-Large-Preview23.2%estimated ± 4.5 pp, medium confidence
188Trinity-Large-Thinking23.2%estimated ± 4.5 pp, medium confidence
189Nemotron 3 Nano 30B20.7%estimated ± 4.5 pp, medium confidence
190DeepSeek V317.0%estimated ± 4.5 pp, medium confidence
191Celeris-114.5%estimated ± 4.5 pp, medium confidence
192Llama 4 Maverick13.9%estimated ± 4.5 pp, medium confidence
193Llama 4 Scout12.5%estimated ± 4.5 pp, medium confidence
194Gemma 3 27B3.5%estimated ± 4.5 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General