Agentic · tools

ResearchClawBench leaderboard

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

As of 2026-10-07, the highest measured score on ResearchClawBench is 21.1% by Claude Opus 4.8. 154 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Ornith-1.5-397B	21.6%	estimated ± 1.3 pp, medium confidence
2	Claude Fable 5.1	21.5%	estimated ± 2.0 pp, medium confidence
3	Muse Spark 1.3	21.4%	estimated ± 2.0 pp, medium confidence
4	GLM-5.3	21.2%	estimated ± 2.0 pp, medium confidence
5	Grok 4.6	21.2%	estimated ± 2.0 pp, medium confidence
6	GPT-6 Astra	21.1%	estimated ± 2.0 pp, medium confidence
7	Claude Opus 4.8	21.1%	measured
8	Claude Fable 5	21.1%	estimated ± 2.0 pp, medium confidence
9	GPT-5.6 Sol	21.0%	estimated ± 2.0 pp, medium confidence
10	Qwen3.8 Max Preview	21.0%	estimated ± 2.0 pp, medium confidence
11	K-EXAONE 2.0	20.8%	estimated ± 1.3 pp, medium confidence
12	Qwen3.8-27B	20.7%	estimated ± 2.0 pp, medium confidence
13	Claude Opus 4.7	20.7%	measured
14	GLM-5.2	20.7%	measured
15	Ornith-1.0-397B	20.6%	estimated ± 1.3 pp, medium confidence
16	Claude Opus 5.5	20.6%	estimated ± 2.2 pp, low confidence
17	Claude Sonnet 5.5	20.5%	estimated ± 2.2 pp, low confidence
18	Claude Sonnet 5	20.5%	estimated ± 2.0 pp, medium confidence
19	Muse Spark 1.2	20.5%	estimated ± 2.0 pp, medium confidence
20	GPT-5.6 Terra	20.5%	estimated ± 2.0 pp, medium confidence
21	GPT-5.6 Luna	20.3%	estimated ± 2.0 pp, medium confidence
22	Grok 4.5	20.3%	estimated ± 2.0 pp, high confidence
23	Grok 4.7	20.2%	estimated ± 2.2 pp, low confidence
24	Gemini 3.8 Flash	20.2%	estimated ± 2.0 pp, high confidence
25	MiMo-V2.6-Pro	20.1%	estimated ± 2.2 pp, low confidence
26	Atria Dawn Preview	20.0%	estimated ± 2.2 pp, low confidence
27	Qwen3.8-Flash-Next	19.9%	estimated ± 2.2 pp, low confidence
28	Gemini 4 Argon	19.9%	estimated ± 2.2 pp, low confidence
29	Ling 3.1 Flash	19.9%	estimated ± 2.2 pp, low confidence
30	Claude Opus 4.6	19.9%	measured
31	Muse Spark 1.1	19.9%	estimated ± 1.8 pp, low confidence
32	MiMo-V2.6-Flash	19.9%	estimated ± 2.2 pp, low confidence
33	DeepSeek V4.1 Flash	19.8%	estimated ± 2.2 pp, low confidence
34	MiniMax M3	19.8%	measured
35	GPT-6.1 Sol	19.8%	estimated ± 2.2 pp, low confidence
36	dots3-note Preview	19.7%	estimated ± 1.3 pp, high confidence
37	Claude Opus 5	19.7%	estimated ± 1.8 pp, low confidence
38	GPT-5.5 Pro	19.6%	estimated ± 2.2 pp, low confidence
39	Step 5 Preview	19.6%	estimated ± 1.8 pp, low confidence
40	Gemini 3.7 Flash	19.6%	estimated ± 2.0 pp, high confidence
41	GPT-5.4 Pro	19.5%	estimated ± 2.2 pp, low confidence
42	GPT-6 Sol	19.5%	estimated ± 2.2 pp, low confidence
43	Kimi K3	19.5%	estimated ± 1.8 pp, low confidence
44	Ornith-1.5-35B-A3B	19.5%	estimated ± 1.3 pp, high confidence
45	Qwen3.6-27B	19.5%	estimated ± 1.3 pp, high confidence
46	Hy4 preview	19.5%	estimated ± 1.8 pp, low confidence
47	Claude Mythos 5	19.3%	estimated ± 2.2 pp, low confidence
48	GPT-6 Luna	19.2%	estimated ± 2.2 pp, medium confidence
49	Mistral Large 4	19.2%	estimated ± 2.2 pp, medium confidence
50	Quasar 438B	19.1%	estimated ± 2.0 pp, high confidence
51	Inkling-Small	19.1%	estimated ± 1.8 pp, medium confidence
52	Beam	19.0%	estimated ± 1.8 pp, medium confidence
53	Claude Opus 4.7 (Adaptive)	18.8%	estimated ± 1.8 pp, medium confidence
54	Ornith-1.0-35B	18.8%	estimated ± 1.3 pp, high confidence
55	Kimi K2.7 Code	18.7%	estimated ± 1.8 pp, medium confidence
56	Gemini 3.6 Flash	18.7%	estimated ± 2.0 pp, high confidence
57	Qwen3.7 Max	18.7%	measured
58	Muse Glimmer 30B	18.7%	estimated ± 1.8 pp, medium confidence
59	Qwen3.6-35B-A3B	18.6%	estimated ± 1.3 pp, high confidence
60	Inkling	18.5%	estimated ± 1.8 pp, medium confidence
61	DeepSeek V4 Pro 0813	18.5%	estimated ± 1.8 pp, medium confidence
62	Claude Sonnet 4.6	18.3%	estimated ± 1.3 pp, high confidence
63	GLM-5.1	18.2%	measured
64	Step 3.7 Flash	18.2%	estimated ± 1.3 pp, high confidence
65	DeepSeek V4 Flash 0731	18.0%	estimated ± 1.8 pp, medium confidence
66	Ornith-1.5-9B	18.0%	estimated ± 1.3 pp, high confidence
67	Gemini 3.5 Flash	18.0%	measured
68	Kimi K2.6	18.0%	measured
69	Qwen3.6 Plus	18.0%	measured
70	Gemini 3 Pro	18.0%	estimated ± 2.2 pp, medium confidence
71	Apodex 1.1	17.9%	estimated ± 2.2 pp, medium confidence
72	Apodex 1.1 Mini	17.9%	estimated ± 2.2 pp, medium confidence
73	Hy3	17.9%	estimated ± 2.0 pp, high confidence
74	Hy3 Preview	17.9%	estimated ± 2.0 pp, high confidence
75	Ling 3.0 Flash	17.7%	estimated ± 1.8 pp, medium confidence
76	Ling 3.0 Flash VL	17.7%	estimated ± 2.2 pp, medium confidence
77	LLaDA2.2-flash	17.5%	estimated ± 1.3 pp, high confidence
78	MiMo-V2.5-Pro	17.4%	estimated ± 1.3 pp, high confidence
79	Muse Spark	17.4%	estimated ± 1.3 pp, high confidence
80	Agents-A1	17.3%	estimated ± 2.2 pp, medium confidence
81	Solar Pro 4	17.3%	estimated ± 1.8 pp, medium confidence
82	Ornith-1.0-9B	17.2%	estimated ± 1.3 pp, high confidence
83	LFM2.5-2.6B	17.1%	estimated ± 1.3 pp, high confidence
84	Qwen3.7 Plus	17.1%	estimated ± 1.3 pp, high confidence
85	GPT-5.3 Codex	17.0%	estimated ± 2.2 pp, medium confidence
86	Solar Open 2	17.0%	estimated ± 1.8 pp, medium confidence
87	GPT-5.5	17.0%	measured
88	GPT-5.4 mini	17.0%	estimated ± 1.8 pp, medium confidence
89	Nemotron 3 Ultra	16.9%	estimated ± 2.0 pp, high confidence
90	MiMo-V2.5	16.9%	measured
91	GPT-5.4 nano	16.8%	estimated ± 1.8 pp, medium confidence
92	Ling 3.0 Flash FP8	16.7%	estimated ± 2.0 pp, high confidence
93	Claude Opus 4.6 (Adaptive)	16.5%	estimated ± 2.6 pp, medium confidence
94	Claude Opus 4.5	16.3%	estimated ± 1.3 pp, high confidence
95	GPT-5.2-Codex	16.1%	estimated ± 2.2 pp, medium confidence
96	A.X K2	15.8%	estimated ± 2.2 pp, medium confidence
97	GLM-5	15.8%	estimated ± 1.3 pp, high confidence
98	Agents-A1-4B	15.8%	estimated ± 2.2 pp, medium confidence
99	GPT-5.1-Codex	15.8%	estimated ± 2.2 pp, medium confidence
100	LongCat-Flash-Lite-Sparse	15.8%	estimated ± 1.8 pp, medium confidence
101	Grok Build 0.1	15.7%	estimated ± 2.2 pp, medium confidence
102	LLaDA2.2-mini	15.7%	estimated ± 1.3 pp, high confidence
103	GPT-5.2	15.6%	estimated ± 2.2 pp, medium confidence
104	Claude Sonnet 4.5	15.6%	estimated ± 2.2 pp, medium confidence
105	Grok 4.1 Fast	15.4%	estimated ± 2.2 pp, medium confidence
106	GLM-5-Turbo	15.4%	estimated ± 1.3 pp, high confidence
107	GPT-5.4	15.3%	measured
108	MiMo-V2-Pro	15.3%	measured
109	GPT-5 (high)	15.3%	estimated ± 2.2 pp, medium confidence
110	Gemini 3.5 Flash-Lite	15.0%	estimated ± 2.0 pp, high confidence
111	Qwen3 Max	14.9%	estimated ± 2.2 pp, medium confidence
112	GLM-5V-Turbo	14.8%	estimated ± 1.3 pp, high confidence
113	Qwen3.5-27B	14.8%	estimated ± 2.2 pp, medium confidence
114	Qwen3.5-35B-A3B	14.8%	estimated ± 2.2 pp, medium confidence
115	Kimi K2.5 (Reasoning)	14.7%	estimated ± 2.2 pp, medium confidence
116	Grok 4	14.7%	estimated ± 2.2 pp, medium confidence
117	Claude 4 Sonnet	14.2%	estimated ± 2.2 pp, medium confidence
118	Qwen3.5 397B	14.2%	measured
119	Gemini 3.1 Flash-Lite	14.0%	estimated ± 2.2 pp, medium confidence
120	Grok 4.20	14.0%	estimated ± 2.2 pp, medium confidence
121	Kimi K2.5	14.0%	measured
122	GPT-5.1	13.9%	estimated ± 2.2 pp, medium confidence
123	Gemini 3 Flash	13.7%	estimated ± 1.3 pp, medium confidence
124	MiniMax M2.7	13.5%	estimated ± 1.3 pp, medium confidence
125	Grok 4.1	13.5%	measured
126	Gemini 3.1 Pro	13.3%	measured
127	GLM-4.7	13.0%	estimated ± 2.2 pp, low confidence
128	MiMo-V2-Omni	12.6%	estimated ± 1.3 pp, medium confidence
129	Grok 4.3	12.4%	measured
130	GPT-4.1	12.0%	estimated ± 2.2 pp, low confidence
131	Qwen3.5-122B-A10B	11.8%	estimated ± 2.0 pp, medium confidence
132	Mistral Medium 3.5 128B	11.6%	estimated ± 2.0 pp, medium confidence
133	DeepSeek V3.2	11.3%	estimated ± 1.3 pp, medium confidence
134	MiniCPM5-2B	11.2%	estimated ± 2.2 pp, low confidence
135	Gemma 4 31B	9.6%	estimated ± 2.0 pp, medium confidence
136	GPT-OSS 120B	9.1%	estimated ± 2.0 pp, medium confidence
137	Nemotron 3.5 Lightning 30B A3B NVFP4	9.0%	estimated ± 2.0 pp, medium confidence
138	MiMo-V2-Flash	8.5%	estimated ± 2.2 pp, low confidence
139	Granite 4.2 8B	6.3%	estimated ± 2.0 pp, medium confidence
140	Command A+	6.2%	estimated ± 2.0 pp, medium confidence
141	Gemini 2.5 Pro	6.1%	estimated ± 2.0 pp, medium confidence
142	Granite 4.2 30B	5.8%	estimated ± 2.2 pp, low confidence
143	Gemma 4 26B A4B	5.2%	estimated ± 2.2 pp, low confidence
144	Ling 3.0 Tiny	5.0%	estimated ± 2.2 pp, low confidence
145	Mistral Large 3	4.6%	estimated ± 2.0 pp, medium confidence
146	Mistral Small 4	2.9%	estimated ± 2.0 pp, medium confidence
147	Mistral Small 4 (Reasoning)	2.9%	estimated ± 2.0 pp, medium confidence
148	GPT-OSS 20B	2.8%	estimated ± 2.0 pp, medium confidence
149	Trinity-Large-Preview	2.4%	estimated ± 2.0 pp, medium confidence
150	Trinity-Large-Thinking	2.4%	estimated ± 2.0 pp, medium confidence
151	Nemotron 3 Nano 30B	2.1%	estimated ± 2.0 pp, medium confidence
152	DeepSeek V3	1.7%	estimated ± 2.0 pp, medium confidence
153	Nemotron 3 Super 100B	1.6%	estimated ± 1.3 pp, medium confidence
154	Celeris-1	1.4%	estimated ± 2.0 pp, medium confidence
155	Llama 4 Maverick	1.3%	estimated ± 2.0 pp, medium confidence
156	Llama 4 Scout	1.2%	estimated ± 2.0 pp, medium confidence
157	Gemma 3 27B	0.3%	estimated ± 2.0 pp, medium confidence
158	DeepSeek V3 0324	0.0%	estimated ± 2.2 pp, low confidence
159	Gemma 4 12B	0.0%	estimated ± 2.2 pp, low confidence
160	Gemma 4 E2B	0.0%	estimated ± 2.2 pp, low confidence
161	Gemma 4 E4B	0.0%	estimated ± 2.2 pp, low confidence
162	GPT-4.1 mini	0.0%	estimated ± 2.2 pp, low confidence
163	GPT-4.1 nano	0.0%	estimated ± 2.2 pp, low confidence
164	GPT-4o	0.0%	estimated ± 2.2 pp, low confidence
165	GPT-4o mini	0.0%	estimated ± 2.2 pp, low confidence
166	Granite 4.2 3B	0.0%	estimated ± 2.2 pp, low confidence
167	K-Exaone	0.0%	estimated ± 2.2 pp, low confidence
168	Ling 2.6 Flash	0.0%	estimated ± 2.2 pp, low confidence
169	Mercury 2.5	0.0%	estimated ± 2.2 pp, low confidence
170	Nemotron 3 Nano Omni 30B A3B	0.0%	estimated ± 2.2 pp, low confidence
171	North Mini Code	0.0%	estimated ± 2.2 pp, low confidence
172	Solar Pro 3	0.0%	estimated ± 2.2 pp, low confidence
173	Ultravox v0.6 Llama 3.3 70B	0.0%	estimated ± 2.2 pp, low confidence

ResearchClawBench leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General