benchgap
Agentic · tools

Claw-Eval leaderboard

As of 2026-11-21, the highest measured score on Claw-Eval is 81.4% by Ornith-1.5-397B. 101 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Muse Wonder Full83.8%estimated ± 4.5 pp, low confidence
2Ornith-1.5-397B81.4%measured
3Inkling 2 Small81.4%estimated ± 4.5 pp, low confidence
4Muse Wonder81.4%estimated ± 4.5 pp, low confidence
5GPT-5.6 Sol80.0%estimated ± 6.4 pp, low confidence
6Kimi K3.1 Pro80.0%estimated ± 6.4 pp, low confidence
7GPT-6 Astra80.0%estimated ± 6.4 pp, medium confidence
8GPT-5.5 Pro80.0%estimated ± 6.4 pp, medium confidence
9GPT-5.4 Pro80.0%estimated ± 6.4 pp, medium confidence
10GPT-5.6 Terra80.0%estimated ± 6.4 pp, medium confidence
11Muse Spark 1.378.6%estimated ± 4.5 pp, medium confidence
12LongCat-2.578.3%measured
13K-EXAONE 2.077.7%measured
14Ultramarine 1 Flash77.6%estimated ± 4.5 pp, medium confidence
15Mistral Large 477.4%estimated ± 6.4 pp, medium confidence
16MiMo-V2.6-Pro77.1%estimated ± 4.5 pp, medium confidence
17Ornith-1.0-397B77.1%measured
18Muse Spark Open77.0%estimated ± 4.5 pp, medium confidence
19Hy4 preview77.0%estimated ± 4.5 pp, medium confidence
20MiMo-V2.6-Flash76.7%estimated ± 4.5 pp, medium confidence
21Step 5 Preview75.6%estimated ± 4.5 pp, medium confidence
22MiniMax M374.5%measured
23Qwen3.8-Flash-Next74.0%estimated ± 4.5 pp, medium confidence
24Claude Opus 573.5%estimated ± 6.3 pp, low confidence
25Muse Spark 1.173.5%estimated ± 4.5 pp, medium confidence
26dots3-note Preview73.4%measured
27GLM-5.273.3%estimated ± 3.2 pp, medium confidence
28ERNIE 5.5 Flash Preview73.1%estimated ± 6.3 pp, low confidence
29Qwen3.8 Max72.8%estimated ± 4.5 pp, medium confidence
30Ling 3.1 Flash72.6%estimated ± 6.3 pp, low confidence
31Kimi K372.6%estimated ± 4.5 pp, medium confidence
32Ornith-1.5-35B-A3B72.5%measured
33Qwen3.6-27B72.4%measured
34Beam 1.0 501B-A23B FP872.2%estimated ± 6.3 pp, low confidence
35Beam 1.0 501B-A23B72.0%estimated ± 6.3 pp, low confidence
36Beam 1.0 501B-A23B NVFP471.9%estimated ± 6.3 pp, low confidence
37Claude Fable 571.9%estimated ± 6.2 pp, low confidence
38Claude Mythos 571.9%estimated ± 6.2 pp, low confidence
39Inkling-Small71.6%estimated ± 6.3 pp, low confidence
40Atria Dawn Preview71.3%estimated ± 4.5 pp, medium confidence
41Inkling 2 Mini70.9%estimated ± 4.5 pp, medium confidence
42Gemini 3.6 Flash70.8%estimated ± 6.2 pp, low confidence
43Kimi K2.7 Code70.6%estimated ± 6.3 pp, low confidence
44Holo3-35B-A3B70.5%estimated ± 6.2 pp, low confidence
45Ling 3.0 Flash Sante70.5%estimated ± 6.3 pp, low confidence
46Claude Opus 4.670.4%measured
47Inkling70.0%estimated ± 6.3 pp, low confidence
48Claude Opus 4.870.0%estimated ± 2.9 pp, medium confidence
49Spark-X2.6-4B69.9%measured
50DeepSeek V4 Pro69.9%estimated ± 6.3 pp, low confidence
51Ornith-1.0-35B69.8%measured
52Claude Sonnet 569.8%estimated ± 6.2 pp, low confidence
53Ripple-65B Preview69.3%estimated ± 6.3 pp, low confidence
54Qwen3.6-35B-A3B68.7%measured
55LFM3-1.7B68.6%measured
56GPT-5.6 Luna68.6%estimated ± 6.4 pp, medium confidence
57DeepSeek V4 Flash 073168.5%estimated ± 6.3 pp, low confidence
58Holo3-122B-A10B68.4%estimated ± 6.2 pp, low confidence
59Claude Opus 4.768.4%estimated ± 2.9 pp, medium confidence
60Claude Sonnet 4.667.8%measured
61Ling 3.0 Flash67.5%estimated ± 6.3 pp, low confidence
62UI-Mate-27B67.4%estimated ± 6.2 pp, low confidence
63Step 3.7 Flash67.1%measured
64Ornith-1.5-9B66.5%measured
65Solar Pro 466.3%estimated ± 6.3 pp, low confidence
66Gemini 3.5 Flash-Lite65.6%estimated ± 6.2 pp, low confidence
67Solar Open 265.3%estimated ± 6.3 pp, low confidence
68Qwen3.7 Max65.2%measured
69GPT-5.4 nano64.7%estimated ± 6.3 pp, low confidence
70Nemotron 3.5 Bolt 74B-A9B BF1664.6%estimated ± 6.3 pp, low confidence
71GPT-5.4 mini64.5%estimated ± 6.2 pp, low confidence
72LLaDA2.2-flash64.2%measured
73GPT-5.563.9%estimated ± 2.9 pp, medium confidence
74Gemini 3.5 Flash63.9%estimated ± 2.9 pp, high confidence
75MiMo-V2.5-Pro63.8%measured
76Muse Spark63.8%measured
77Nemotron 3.5 Bolt 74B-A9B63.6%estimated ± 6.3 pp, low confidence
78GPT-5.263.3%estimated ± 4.5 pp, medium confidence
79Ornith-1.0-9B63.1%measured
80Sage-A1.563.1%estimated ± 6.4 pp, medium confidence
81GPT-5.3 Codex63.0%estimated ± 4.5 pp, medium confidence
82LFM2.5-2.6B62.8%measured
83Qwen3.8-27B62.8%estimated ± 4.5 pp, medium confidence
84Nemotron 3.5 Super VL 120B A12B62.7%estimated ± 6.4 pp, medium confidence
85Nemotron 3.5 Super VL 120B A12B NVFP462.7%estimated ± 6.4 pp, medium confidence
86Agents-A162.7%estimated ± 6.4 pp, medium confidence
87Agents-A1-4B62.7%estimated ± 6.4 pp, medium confidence
88GLM-4.762.7%estimated ± 6.4 pp, low confidence
89Nemotron 3.5 Lightning 30B A3B NVFP462.7%estimated ± 6.4 pp, low confidence
90Nemotron 3 Ultra62.7%estimated ± 6.4 pp, low confidence
91Sarvam 105B62.7%estimated ± 6.4 pp, low confidence
92Sarvam 30B62.7%estimated ± 6.4 pp, low confidence
93Qwen3.7 Plus62.7%measured
94GLM-5.162.3%measured
95Kimi K2.662.3%measured
96MiMo-V2.562.3%measured
97LongCat-Flash-Lite-Sparse61.6%estimated ± 6.3 pp, low confidence
98UI-Mate-9B60.8%estimated ± 6.2 pp, low confidence
99Muse Glimmer 30B60.6%estimated ± 6.2 pp, low confidence
100GPT-5.460.3%measured
101Claude Sonnet 4.560.0%estimated ± 4.5 pp, medium confidence
102Claude Opus 4.559.6%measured
103Grok Build 0.159.4%estimated ± 6.8 pp, low confidence
104GPT-5.1-Codex59.2%estimated ± 4.5 pp, medium confidence
105GPT-5.2-Codex59.1%estimated ± 4.5 pp, medium confidence
106Qwen3.6 Plus58.8%measured
107Grok 4.1 Fast58.7%estimated ± 6.8 pp, low confidence
108Gemini 3.1 Pro57.8%measured
109MiMo-V2-Pro57.8%measured
110GLM-557.7%measured
111LLaDA2.2-mini57.2%measured
112Qwen3 Max57.1%estimated ± 6.8 pp, low confidence
113Claude 4.1 Opus57.0%estimated ± 4.5 pp, medium confidence
114Qwen3.5 397B56.8%measured
115Grok 456.5%estimated ± 6.8 pp, low confidence
116Gemini 2.5 Pro56.3%estimated ± 6.8 pp, low confidence
117GPT-5.155.9%estimated ± 6.8 pp, low confidence
118GLM-5-Turbo55.8%measured
119Grok 4.155.4%estimated ± 3.2 pp, high confidence
120Qwen3.5 Plus55.3%estimated ± 4.5 pp, medium confidence
121Qwen3.5-122B-A10B55.3%estimated ± 6.2 pp, low confidence
122Claude 4 Sonnet55.3%estimated ± 4.5 pp, medium confidence
123Mistral Medium 3.5 128B54.8%estimated ± 6.8 pp, low confidence
124Gemini 3.1 Flash-Lite54.5%estimated ± 6.8 pp, low confidence
125Grok 4.2054.5%estimated ± 6.8 pp, low confidence
126Claude Haiku 4.554.1%estimated ± 4.5 pp, medium confidence
127Qwen3.5-27B54.1%estimated ± 6.2 pp, low confidence
128GLM-5V-Turbo53.8%measured
129Hy3 Preview53.6%estimated ± 6.8 pp, low confidence
130Grok 4.353.1%estimated ± 2.9 pp, medium confidence
131Qwen3.5-35B-A3B52.9%estimated ± 6.2 pp, low confidence
132Gemma 4 31B52.7%estimated ± 6.8 pp, low confidence
133Kimi K2.552.3%measured
134Gemini 3 Pro51.8%estimated ± 4.5 pp, medium confidence
135Trinity-Large-Thinking51.0%estimated ± 6.8 pp, low confidence
136GPT-550.4%estimated ± 4.5 pp, low confidence
137Gemini 3 Flash49.2%measured
138GPT-OSS 120B48.9%estimated ± 6.8 pp, low confidence
139MiniMax M2.748.7%measured
140GPT-4.145.8%estimated ± 6.8 pp, low confidence
141MiMo-V2-Omni45.2%measured
142DeepSeek V3.240.2%measured
143Nemotron 3 Super 100B5.5%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General