Agentic · tools

DeepSearchQA leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on DeepSearchQA is 96.0% by Atria Dawn Preview. 148 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Fugu Cyber	100.0%	estimated ± 11.5 pp, low confidence
2	Atria Dawn Preview	96.0%	measured
3	Claude Opus 5	95.0%	measured
4	Kimi K3	95.0%	measured
5	Claude Fable 5.1	94.3%	estimated ± 4.5 pp, medium confidence
6	GLM-5.3	93.8%	estimated ± 4.5 pp, high confidence
7	Grok 4.6	93.8%	estimated ± 4.5 pp, high confidence
8	GPT-6 Astra	93.6%	estimated ± 4.5 pp, high confidence
9	Claude Fable 5	93.5%	estimated ± 4.5 pp, high confidence
10	GPT-5.6 Sol	93.4%	estimated ± 4.5 pp, high confidence
11	Qwen3.8 Max Preview	93.3%	estimated ± 4.5 pp, high confidence
12	DeepSeek V4 Pro 0813	93.3%	estimated ± 4.5 pp, high confidence
13	Claude Opus 4.8	93.1%	measured
14	Qwen3.8-27B	92.9%	estimated ± 4.5 pp, high confidence
15	Step 3.7 Flash	92.8%	measured
16	Claude Sonnet 5	92.5%	estimated ± 4.5 pp, high confidence
17	Kimi K2.6	92.5%	measured
18	Muse Spark 1.2	92.5%	estimated ± 4.5 pp, high confidence
19	GPT-5.6 Terra	92.4%	estimated ± 4.5 pp, high confidence
20	Apodex 1.1	92.4%	measured
21	GPT-5.6 Luna	92.2%	estimated ± 4.5 pp, high confidence
22	Grok 4.5	92.1%	estimated ± 4.5 pp, high confidence
23	GPT-5.5 Pro	92.1%	estimated ± 8.1 pp, low confidence
24	dots3-note Preview	92.1%	measured
25	DeepSeek V4 Flash 0731	92.1%	estimated ± 4.5 pp, high confidence
26	Gemini 3.8 Flash	91.9%	estimated ± 4.5 pp, high confidence
27	Claude Opus 4.7 (Adaptive)	91.6%	estimated ± 4.5 pp, high confidence
28	GLM-5.2	91.6%	estimated ± 4.5 pp, high confidence
29	GPT-5.4 Pro	91.6%	estimated ± 8.1 pp, low confidence
30	GPT-5.5	91.1%	estimated ± 4.5 pp, high confidence
31	Step 5 Preview	91.1%	estimated ± 8.1 pp, low confidence
32	Gemini 3.7 Flash	90.9%	estimated ± 4.5 pp, high confidence
33	Claude Mythos 5	90.6%	estimated ± 8.1 pp, low confidence
34	Claude Opus 5.5	90.5%	estimated ± 8.4 pp, low confidence
35	Claude Sonnet 5.5	90.5%	estimated ± 8.4 pp, low confidence
36	Grok 4.7	90.5%	estimated ± 8.4 pp, medium confidence
37	MiMo-V2.6-Pro	90.5%	estimated ± 8.4 pp, medium confidence
38	Qwen3.8-Flash-Next	90.4%	estimated ± 8.4 pp, medium confidence
39	Gemini 4 Argon	90.4%	estimated ± 8.4 pp, medium confidence
40	Ling 3.1 Flash	90.4%	estimated ± 8.4 pp, medium confidence
41	MiMo-V2.6-Flash	90.4%	estimated ± 8.4 pp, medium confidence
42	DeepSeek V4.1 Flash	90.4%	estimated ± 8.4 pp, medium confidence
43	GPT-6.1 Sol	90.4%	estimated ± 8.4 pp, medium confidence
44	GPT-6 Sol	90.3%	estimated ± 8.4 pp, medium confidence
45	Hy4 preview	90.2%	estimated ± 10.9 pp, low confidence
46	GPT-6 Luna	90.2%	estimated ± 8.4 pp, medium confidence
47	Mistral Large 4	90.1%	estimated ± 8.4 pp, medium confidence
48	Gemini 3.8 Flash Cyber	90.0%	estimated ± 11.5 pp, low confidence
49	Quasar 438B	89.9%	estimated ± 4.5 pp, high confidence
50	Ornith-1.5-397B	89.6%	estimated ± 8.1 pp, low confidence
51	Muse Spark 1.3	89.4%	measured
52	MiniMax M3	89.4%	estimated ± 4.5 pp, high confidence
53	Gemini 3.6 Flash	89.2%	estimated ± 4.5 pp, high confidence
54	Qwen3.8 Max	88.9%	estimated ± 8.5 pp, low confidence
55	Apodex 1.1 Mini	88.8%	estimated ± 8.4 pp, medium confidence
56	Ling 3.0 Flash VL	88.4%	estimated ± 8.4 pp, medium confidence
57	Gemini 3.5 Flash	88.1%	estimated ± 4.5 pp, high confidence
58	Hy3	87.4%	estimated ± 4.5 pp, high confidence
59	Hy3 Preview	87.4%	estimated ± 4.5 pp, high confidence
60	GLM-5.1	87.2%	estimated ± 4.5 pp, high confidence
61	Inkling-Small	87.1%	estimated ± 4.5 pp, high confidence
62	Inkling	86.8%	estimated ± 4.5 pp, high confidence
63	Holo3-35B-A3B	86.8%	estimated ± 8.5 pp, low confidence
64	Qwen3.7 Max	86.6%	estimated ± 4.5 pp, high confidence
65	MiMo-V2.5-Pro	85.9%	estimated ± 4.5 pp, high confidence
66	Kimi K2.7 Code	85.8%	estimated ± 4.5 pp, high confidence
67	Nemotron 3 Ultra	85.4%	estimated ± 4.5 pp, high confidence
68	Ling 3.0 Flash	84.9%	estimated ± 4.5 pp, high confidence
69	Ling 3.0 Flash FP8	84.9%	estimated ± 4.5 pp, high confidence
70	Muse Spark 1.1	84.9%	measured
71	Holo3-122B-A10B	84.5%	estimated ± 8.5 pp, low confidence
72	Qwen3.6 Plus	84.5%	estimated ± 8.4 pp, medium confidence
73	Qwen3.6-27B	84.3%	estimated ± 4.5 pp, high confidence
74	Qwen3.7 Plus	84.1%	estimated ± 4.5 pp, high confidence
75	GPT-5.4 mini	84.0%	estimated ± 4.5 pp, high confidence
76	UI-Mate-27B	83.4%	estimated ± 8.5 pp, low confidence
77	A.X K2	83.3%	estimated ± 8.4 pp, medium confidence
78	Agents-A1	82.7%	estimated ± 8.1 pp, low confidence
79	GPT-5.4 nano	82.6%	estimated ± 4.5 pp, high confidence
80	Grok 4.3	82.2%	estimated ± 4.5 pp, high confidence
81	MiniMax M2.7	81.8%	estimated ± 4.5 pp, high confidence
82	GPT-5 (high)	81.2%	estimated ± 8.4 pp, medium confidence
83	Gemini 3.5 Flash-Lite	80.9%	estimated ± 4.5 pp, high confidence
84	MiMo-V2.5	80.5%	estimated ± 10.2 pp, low confidence
85	Beam	80.1%	measured
86	Qwen3.6-35B-A3B	80.0%	estimated ± 4.5 pp, high confidence
87	Ornith-1.5-35B-A3B	79.2%	estimated ± 8.1 pp, low confidence
88	Agents-A1-4B	78.9%	estimated ± 8.1 pp, low confidence
89	GPT-5.2	78.6%	estimated ± 8.1 pp, low confidence
90	GPT-5.1-Codex	78.3%	estimated ± 10.9 pp, low confidence
91	GPT-5.2-Codex	78.3%	estimated ± 10.9 pp, low confidence
92	Qwen3.5 397B	77.5%	estimated ± 8.1 pp, low confidence
93	Qwen3.5-27B	77.2%	estimated ± 8.1 pp, low confidence
94	Qwen3.5-35B-A3B	77.2%	estimated ± 8.1 pp, low confidence
95	Claude Opus 4.7	77.2%	estimated ± 7.5 pp, low confidence
96	Kimi K2.5 (Reasoning)	77.1%	estimated ± 8.1 pp, low confidence
97	Kimi K2.5	77.1%	measured
98	MiMo-V2-Pro	77.0%	estimated ± 10.2 pp, low confidence
99	Claude 4.1 Opus	76.9%	estimated ± 10.9 pp, low confidence
100	Gemini 3.5 Flash Cyber	76.8%	estimated ± 11.5 pp, low confidence
101	Claude Mythos Preview	76.8%	estimated ± 11.5 pp, low confidence
102	GLM-5	76.8%	estimated ± 11.5 pp, low confidence
103	Ornith-1.5-9B	76.3%	estimated ± 8.1 pp, low confidence
104	Claude Opus 4.5	76.2%	estimated ± 8.5 pp, low confidence
105	UI-Mate-9B	76.1%	estimated ± 8.5 pp, low confidence
106	Qwen3.5 Plus	75.7%	estimated ± 10.9 pp, low confidence
107	Claude 4 Sonnet	75.7%	estimated ± 10.9 pp, low confidence
108	GLM-4.7	75.7%	estimated ± 8.1 pp, low confidence
109	Solar Pro 4	75.4%	estimated ± 8.1 pp, low confidence
110	LongCat-Flash-Lite-Sparse	75.3%	estimated ± 8.1 pp, low confidence
111	GPT-5.1	75.3%	estimated ± 8.4 pp, medium confidence
112	GPT-5.3 Codex	75.1%	estimated ± 8.5 pp, low confidence
113	Claude Haiku 4.5	74.9%	estimated ± 10.9 pp, low confidence
114	Muse Spark	74.8%	measured
115	Muse Glimmer 30B	74.6%	measured
116	Claude Opus 4.6	73.7%	measured
117	GPT-5.4	73.6%	measured
118	Claude Sonnet 4.6	73.4%	estimated ± 7.5 pp, low confidence
119	Gemini 3 Flash	73.4%	estimated ± 10.9 pp, low confidence
120	Gemini 3 Pro	73.4%	estimated ± 10.9 pp, low confidence
121	Claude Sonnet 4.5	72.6%	estimated ± 8.5 pp, low confidence
122	Grok 4.1	72.6%	estimated ± 10.2 pp, low confidence
123	Qwen3.5-122B-A10B	71.8%	estimated ± 4.5 pp, medium confidence
124	Mistral Medium 3.5 128B	71.1%	estimated ± 4.5 pp, medium confidence
125	Gemini 3.1 Pro	69.7%	measured
126	Gemma 4 31B	63.9%	estimated ± 4.5 pp, medium confidence
127	MiniCPM5-2B	63.2%	estimated ± 8.4 pp, medium confidence
128	Grok 4.20	62.8%	measured
129	GPT-OSS 120B	62.0%	estimated ± 4.5 pp, medium confidence
130	Nemotron 3.5 Lightning 30B A3B NVFP4	61.7%	estimated ± 4.5 pp, medium confidence
131	MiMo-V2-Flash	53.0%	estimated ± 8.4 pp, medium confidence
132	Nemotron 3 Super 100B	51.9%	estimated ± 4.5 pp, medium confidence
133	Granite 4.2 8B	49.2%	estimated ± 4.5 pp, medium confidence
134	Command A+	48.7%	estimated ± 4.5 pp, medium confidence
135	Gemini 2.5 Pro	48.1%	estimated ± 4.5 pp, medium confidence
136	Granite 4.2 30B	45.4%	estimated ± 8.4 pp, medium confidence
137	Gemma 4 26B A4B	44.0%	estimated ± 8.4 pp, medium confidence
138	Ling 3.0 Tiny	43.4%	estimated ± 8.4 pp, medium confidence
139	Mistral Large 3	38.9%	estimated ± 4.5 pp, medium confidence
140	DeepSeek V3 0324	34.5%	estimated ± 8.4 pp, medium confidence
141	Gemma 4 12B	34.5%	estimated ± 8.4 pp, medium confidence
142	Gemma 4 E2B	34.5%	estimated ± 8.4 pp, medium confidence
143	Gemma 4 E4B	34.5%	estimated ± 8.4 pp, medium confidence
144	GPT-4.1 mini	34.5%	estimated ± 8.4 pp, medium confidence
145	GPT-4.1 nano	34.5%	estimated ± 8.4 pp, medium confidence
146	GPT-4o	34.5%	estimated ± 8.4 pp, medium confidence
147	GPT-4o mini	34.5%	estimated ± 8.4 pp, medium confidence
148	Granite 4.2 3B	34.5%	estimated ± 8.4 pp, medium confidence
149	K-Exaone	34.5%	estimated ± 8.4 pp, medium confidence
150	LFM2.5-2.6B	34.5%	estimated ± 8.4 pp, medium confidence
151	Ling 2.6 Flash	34.5%	estimated ± 8.4 pp, medium confidence
152	Nemotron 3 Nano Omni 30B A3B	34.5%	estimated ± 8.4 pp, medium confidence
153	North Mini Code	34.5%	estimated ± 8.4 pp, medium confidence
154	Solar Pro 3	34.5%	estimated ± 8.4 pp, medium confidence
155	Ultravox v0.6 Llama 3.3 70B	34.5%	estimated ± 8.4 pp, medium confidence
156	Mercury 2.5	34.0%	measured
157	Mistral Small 4	27.3%	estimated ± 4.5 pp, medium confidence
158	Mistral Small 4 (Reasoning)	27.3%	estimated ± 4.5 pp, medium confidence
159	GPT-OSS 20B	26.7%	estimated ± 4.5 pp, medium confidence
160	Trinity-Large-Preview	23.2%	estimated ± 4.5 pp, medium confidence
161	Trinity-Large-Thinking	23.2%	estimated ± 4.5 pp, medium confidence
162	Nemotron 3 Nano 30B	20.7%	estimated ± 4.5 pp, medium confidence
163	DeepSeek V3	17.0%	estimated ± 4.5 pp, medium confidence
164	Celeris-1	14.5%	estimated ± 4.5 pp, medium confidence
165	Llama 4 Maverick	13.9%	estimated ± 4.5 pp, medium confidence
166	Llama 4 Scout	12.5%	estimated ± 4.5 pp, medium confidence
167	Gemma 3 27B	3.5%	estimated ± 4.5 pp, medium confidence

DeepSearchQA leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General