benchgap
Agentic · tools

ResearchClawBench leaderboard

As of 2026-11-21, the highest measured score on ResearchClawBench is 21.1% by Claude Opus 4.8. 186 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Opus 5.524.2%estimated ± 2.2 pp, low confidence
2Muse Wonder Full24.1%estimated ± 2.2 pp, low confidence
3Kimi K3.1 Pro24.0%estimated ± 2.2 pp, low confidence
4Claude Sonnet 5.523.9%estimated ± 2.2 pp, low confidence
5Ornith-1.5-397B23.8%estimated ± 1.3 pp, medium confidence
6Muse Wonder23.3%estimated ± 2.2 pp, low confidence
7LongCat-2.522.8%estimated ± 1.3 pp, medium confidence
8Grok 4.722.7%estimated ± 2.2 pp, low confidence
9K-EXAONE 2.022.5%estimated ± 1.3 pp, medium confidence
10Ornith-1.0-397B22.3%estimated ± 1.3 pp, medium confidence
11Gemini 4 Argon21.9%estimated ± 2.2 pp, low confidence
12Claude Haiku 5.521.8%estimated ± 2.2 pp, low confidence
13Claude Fable 5.121.7%estimated ± 1.9 pp, medium confidence
14DeepSeek V4.1 Flash 101721.7%estimated ± 2.2 pp, low confidence
15GPT-6.1 Sol21.6%estimated ± 2.2 pp, low confidence
16Muse Spark 1.321.6%estimated ± 1.9 pp, medium confidence
17GLM-5.321.4%estimated ± 1.9 pp, medium confidence
18Grok 4.621.4%estimated ± 1.9 pp, medium confidence
19GPT-6 Astra21.3%estimated ± 1.9 pp, medium confidence
20Claude Fable 521.2%estimated ± 1.9 pp, medium confidence
21GPT-5.6 Sol21.2%estimated ± 1.9 pp, medium confidence
22Ultramarine 1 Flash21.2%estimated ± 2.6 pp, low confidence
23Qwen3.8 Max Preview21.1%estimated ± 1.9 pp, medium confidence
24Claude Opus 4.821.1%measured
25dots3-note Preview21.0%estimated ± 1.3 pp, medium confidence
26Qwen3.8-27B20.9%estimated ± 1.9 pp, medium confidence
27GPT-6 Sol20.8%estimated ± 2.2 pp, low confidence
28Ornith-1.5-35B-A3B20.7%estimated ± 1.3 pp, medium confidence
29Claude Opus 4.720.7%measured
30GLM-5.220.7%measured
31Qwen3.6-27B20.7%estimated ± 1.3 pp, medium confidence
32Qwen3.8-Flash-Next20.7%estimated ± 1.9 pp, medium confidence
33Qwen3.8 Max20.7%estimated ± 1.9 pp, medium confidence
34Claude Sonnet 520.7%estimated ± 1.9 pp, medium confidence
35Muse Spark 1.220.6%estimated ± 1.9 pp, medium confidence
36GPT-5.6 Terra20.6%estimated ± 1.9 pp, medium confidence
37GPT-5.6 Luna20.5%estimated ± 1.9 pp, medium confidence
38Grok 4.520.4%estimated ± 1.9 pp, high confidence
39Atria Dawn Preview20.4%estimated ± 2.3 pp, low confidence
40Claude Mythos 520.3%estimated ± 2.1 pp, low confidence
41Gemini 3.8 Flash20.3%estimated ± 1.9 pp, high confidence
42Qwen4 Max20.3%estimated ± 2.4 pp, low confidence
43MiniMax M3.1 Flash20.3%estimated ± 2.2 pp, low confidence
44Muse Spark Open20.3%estimated ± 1.8 pp, medium confidence
45Muse Spark 1.120.1%estimated ± 1.8 pp, medium confidence
46GPT-6 Luna20.0%estimated ± 2.2 pp, medium confidence
47GPT-5.5 Pro20.0%estimated ± 2.3 pp, low confidence
48Mistral Large 419.9%estimated ± 2.2 pp, medium confidence
49Claude Opus 519.9%estimated ± 1.8 pp, medium confidence
50Claude Opus 4.619.9%measured
51Step 5 Preview19.9%estimated ± 1.8 pp, medium confidence
52GPT-5.4 Pro19.9%estimated ± 2.3 pp, low confidence
53Inkling 2 Small19.8%estimated ± 1.8 pp, medium confidence
54Holo3-35B-A3B19.8%estimated ± 2.1 pp, medium confidence
55Spark-X2.6-4B19.8%estimated ± 1.3 pp, high confidence
56MiniMax M319.8%measured
57ERNIE 5.5 Flash Preview19.8%estimated ± 1.8 pp, medium confidence
58Ornith-1.0-35B19.8%estimated ± 1.3 pp, high confidence
59Gemini 3.7 Flash19.8%estimated ± 1.9 pp, high confidence
60Kimi K319.8%estimated ± 1.8 pp, medium confidence
61Hy4 preview19.7%estimated ± 1.8 pp, medium confidence
62MiMo-V2.6-Pro19.7%estimated ± 2.1 pp, medium confidence
63Ling 3.1 Flash19.6%estimated ± 1.8 pp, high confidence
64Inkling 2 Mini19.6%estimated ± 1.8 pp, high confidence
65Beam 1.0 501B-A23B FP819.5%estimated ± 1.8 pp, high confidence
66MiMo-V2.6-Flash19.4%estimated ± 2.1 pp, medium confidence
67Beam 1.0 501B-A23B19.4%estimated ± 1.8 pp, high confidence
68Qwen3.6-35B-A3B19.4%estimated ± 1.3 pp, high confidence
69Beam 1.0 501B-A23B NVFP419.4%estimated ± 1.8 pp, high confidence
70LFM3-1.7B19.4%estimated ± 1.3 pp, high confidence
71Inkling-Small19.3%estimated ± 1.8 pp, high confidence
72Quasar 438B19.2%estimated ± 1.9 pp, high confidence
73Claude Sonnet 4.619.1%estimated ± 1.3 pp, high confidence
74Holo3-122B-A10B19.0%estimated ± 2.1 pp, medium confidence
75Kimi K2.7 Code18.9%estimated ± 1.8 pp, high confidence
76Ling 3.0 Flash Sante18.9%estimated ± 1.8 pp, high confidence
77Muse Glimmer 30B18.9%estimated ± 1.8 pp, high confidence
78Step 3.7 Flash18.8%estimated ± 1.3 pp, high confidence
79Gemini 3.6 Flash18.8%estimated ± 1.9 pp, high confidence
80Inkling18.7%estimated ± 1.8 pp, high confidence
81Qwen3.7 Max18.7%measured
82DeepSeek V4 Pro18.7%estimated ± 1.8 pp, high confidence
83Ornith-1.5-9B18.6%estimated ± 1.3 pp, high confidence
84UI-Mate-27B18.6%estimated ± 2.1 pp, medium confidence
85Sage-A1.518.5%estimated ± 2.3 pp, medium confidence
86Ripple-65B Preview18.5%estimated ± 1.8 pp, high confidence
87DeepSeek V4 Flash 073118.2%estimated ± 1.8 pp, high confidence
88GLM-5.118.2%measured
89Gemini 3.5 Flash18.0%measured
90Kimi K2.618.0%measured
91Qwen3.6 Plus18.0%measured
92Gemini 3 Pro18.0%estimated ± 2.2 pp, medium confidence
93Hy317.9%estimated ± 1.9 pp, high confidence
94Hy3 Preview17.9%estimated ± 1.9 pp, high confidence
95Ling 3.0 Flash17.8%estimated ± 1.8 pp, high confidence
96LLaDA2.2-flash17.8%estimated ± 1.3 pp, high confidence
97Apodex 1.117.7%estimated ± 2.2 pp, medium confidence
98Apodex 1.1 Mini17.7%estimated ± 2.2 pp, medium confidence
99MiMo-V2.5-Pro17.7%estimated ± 1.3 pp, high confidence
100Muse Spark17.7%estimated ± 1.3 pp, high confidence
101Nemotron 3.5 Super VL 120B A12B17.5%estimated ± 2.2 pp, medium confidence
102Nemotron 3.5 Super VL 120B A12B NVFP417.5%estimated ± 2.2 pp, medium confidence
103Ling 3.0 Flash VL17.4%estimated ± 2.2 pp, medium confidence
104Solar Pro 417.4%estimated ± 1.8 pp, high confidence
105Agents-A117.4%estimated ± 2.3 pp, medium confidence
106Ornith-1.0-9B17.4%estimated ± 1.3 pp, high confidence
107LFM2.5-2.6B17.3%estimated ± 1.3 pp, high confidence
108Qwen3.7 Plus17.3%estimated ± 1.3 pp, high confidence
109Solar Open 217.1%estimated ± 1.8 pp, high confidence
110GPT-5.4 mini17.1%estimated ± 1.8 pp, high confidence
111GPT-5.517.0%measured
112Nemotron 3 Ultra16.9%estimated ± 1.9 pp, high confidence
113MiMo-V2.516.9%measured
114GPT-5.4 nano16.9%estimated ± 1.8 pp, high confidence
115Nemotron 3.5 Bolt 74B-A9B BF1616.8%estimated ± 1.8 pp, high confidence
116Claude Opus 4.6 (Adaptive)16.5%estimated ± 2.4 pp, medium confidence
117Nemotron 3.5 Bolt 74B-A9B16.5%estimated ± 1.8 pp, high confidence
118UI-Mate-9B16.2%estimated ± 2.1 pp, medium confidence
119Claude Opus 4.516.2%estimated ± 1.3 pp, high confidence
120GPT-5.2-Codex16.1%estimated ± 2.2 pp, medium confidence
121GLM-4.716.0%estimated ± 2.2 pp, medium confidence
122GPT-5.3 Codex15.9%estimated ± 2.1 pp, medium confidence
123LongCat-Flash-Lite-Sparse15.8%estimated ± 1.8 pp, high confidence
124GPT-5.1-Codex15.8%estimated ± 2.2 pp, medium confidence
125Agents-A1-4B15.8%estimated ± 2.3 pp, medium confidence
126Grok Build 0.115.7%estimated ± 2.2 pp, medium confidence
127Ling 3.0 Flash Fin15.7%estimated ± 2.4 pp, medium confidence
128Claude 4.1 Opus15.7%estimated ± 2.6 pp, low confidence
129A.X K215.5%estimated ± 2.2 pp, medium confidence
130GLM-515.5%estimated ± 1.3 pp, high confidence
131Grok 4.1 Fast15.4%estimated ± 2.2 pp, medium confidence
132LLaDA2.2-mini15.3%estimated ± 1.3 pp, high confidence
133GPT-5.415.3%measured
134MiMo-V2-Pro15.3%measured
135Mercury 315.3%estimated ± 2.9 pp, low confidence
136Qwen3.5 Plus15.2%estimated ± 2.6 pp, low confidence
137GPT-515.1%estimated ± 2.2 pp, medium confidence
138Claude Sonnet 4.515.1%estimated ± 2.1 pp, low confidence
139Gemini 3.5 Flash-Lite15.0%estimated ± 1.9 pp, high confidence
140Claude Haiku 4.514.9%estimated ± 2.6 pp, low confidence
141Qwen3 Max14.9%estimated ± 2.2 pp, medium confidence
142GLM-5-Turbo14.9%estimated ± 1.3 pp, high confidence
143Grok 414.7%estimated ± 2.2 pp, medium confidence
144GPT-5.114.2%estimated ± 2.2 pp, medium confidence
145Claude 4 Sonnet14.2%estimated ± 2.2 pp, medium confidence
146Qwen3.5 397B14.2%measured
147GLM-5V-Turbo14.1%estimated ± 1.3 pp, high confidence
148Gemini 3.1 Flash-Lite14.0%estimated ± 2.2 pp, medium confidence
149Grok 4.2014.0%estimated ± 2.2 pp, medium confidence
150Kimi K2.514.0%measured
151Qwen3.5-27B13.9%estimated ± 2.1 pp, low confidence
152Qwen3.5-35B-A3B13.5%estimated ± 2.1 pp, low confidence
153Grok 4.113.5%measured
154Gemini 3.1 Pro13.3%measured
155MiniCPM5-2B13.1%estimated ± 2.2 pp, low confidence
156Gemini 3 Flash12.5%estimated ± 1.3 pp, medium confidence
157Grok 4.312.4%measured
158MiniMax M2.712.4%estimated ± 1.3 pp, medium confidence
159MiMo-V2-Flash12.4%estimated ± 2.2 pp, low confidence
160Sarvam 105B12.3%estimated ± 2.3 pp, low confidence
161GPT-4.112.0%estimated ± 2.2 pp, low confidence
162GPT-5.211.9%estimated ± 2.1 pp, low confidence
163Granite 4.2 30B11.8%estimated ± 2.2 pp, low confidence
164Gemma 4 26B A4B11.7%estimated ± 2.2 pp, low confidence
165Qwen3.5-122B-A10B11.7%estimated ± 1.9 pp, medium confidence
166Ling 3.0 Tiny11.7%estimated ± 2.2 pp, low confidence
167Mistral Medium 3.5 128B11.5%estimated ± 1.9 pp, medium confidence
168MiMo-V2-Omni11.1%estimated ± 1.3 pp, medium confidence
169Claude 3.5 Haiku11.1%estimated ± 2.2 pp, low confidence
170DeepSeek V3 032411.1%estimated ± 2.2 pp, low confidence
171Gemma 4 12B11.1%estimated ± 2.2 pp, low confidence
172Gemma 4 E2B11.1%estimated ± 2.2 pp, low confidence
173Gemma 4 E4B11.1%estimated ± 2.2 pp, low confidence
174GPT-4.1 mini11.1%estimated ± 2.2 pp, low confidence
175GPT-4.1 nano11.1%estimated ± 2.2 pp, low confidence
176GPT-4o11.1%estimated ± 2.2 pp, low confidence
177GPT-4o mini11.1%estimated ± 2.2 pp, low confidence
178Granite 4.2 3B11.1%estimated ± 2.2 pp, low confidence
179K-Exaone11.1%estimated ± 2.2 pp, low confidence
180Ling 2.6 Flash11.1%estimated ± 2.2 pp, low confidence
181Llama 3.3 70B11.1%estimated ± 2.2 pp, low confidence
182Mercury 2.511.1%estimated ± 2.2 pp, low confidence
183Mistral Small 3.111.1%estimated ± 2.2 pp, low confidence
184Nemotron 3 Nano Omni 30B A3B11.1%estimated ± 2.2 pp, low confidence
185North Mini Code11.1%estimated ± 2.2 pp, low confidence
186Solar Pro 311.1%estimated ± 2.2 pp, low confidence
187Ultravox v0.6 Llama 3.3 70B11.1%estimated ± 2.2 pp, low confidence
188Gemma 4 31B9.5%estimated ± 1.9 pp, medium confidence
189DeepSeek V3.29.4%estimated ± 1.3 pp, medium confidence
190GPT-OSS 120B9.0%estimated ± 1.9 pp, medium confidence
191Nemotron 3.5 Lightning 30B A3B NVFP48.9%estimated ± 1.9 pp, medium confidence
192Granite 4.2 8B6.2%estimated ± 1.9 pp, medium confidence
193Command A+6.1%estimated ± 1.9 pp, medium confidence
194Gemini 2.5 Pro6.0%estimated ± 1.9 pp, medium confidence
195Mistral Large 34.5%estimated ± 1.9 pp, medium confidence
196Mistral Small 42.8%estimated ± 1.9 pp, medium confidence
197GPT-OSS 20B2.8%estimated ± 1.9 pp, medium confidence
198Trinity-Large-Preview2.3%estimated ± 1.9 pp, medium confidence
199Trinity-Large-Thinking2.3%estimated ± 1.9 pp, medium confidence
200Nemotron 3 Nano 30B2.0%estimated ± 1.9 pp, medium confidence
201DeepSeek V31.6%estimated ± 1.9 pp, medium confidence
202Celeris-11.4%estimated ± 1.9 pp, medium confidence
203Llama 4 Maverick1.3%estimated ± 1.9 pp, medium confidence
204Llama 4 Scout1.2%estimated ± 1.9 pp, medium confidence
205Gemma 3 27B0.3%estimated ± 1.9 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General