benchgap
Agentic · tools

CWE-bench v1 leaderboard

As of 2026-11-19, the highest measured score on CWE-bench v1 is 71.0% by Gemini 4 Carbon. 140 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Gemini 4 Carbon71.0%measured
2GPT-6 Cyber70.8%estimated ± 4.5 pp, medium confidence
3Claude Fable 5.269.3%estimated ± 5.1 pp, medium confidence
4GPT-6.2 Sol69.3%estimated ± 5.1 pp, medium confidence
5GPT-6.2 Astra69.3%estimated ± 5.1 pp, medium confidence
6Muse Wonder Full69.3%estimated ± 5.1 pp, medium confidence
7GPT-6.1 Terra69.3%estimated ± 5.1 pp, medium confidence
8Gemini 4 Argon68.0%measured
9GPT-6 Astra68.0%measured
10Grok 4.768.0%measured
11Claude Opus 5.567.0%measured
12Hy466.4%estimated ± 5.9 pp, low confidence
13Claude Mythos 5.265.4%estimated ± 4.5 pp, medium confidence
14GPT-6.1 Sol63.9%estimated ± 4.5 pp, medium confidence
15GPT-5.6 Sol63.2%estimated ± 4.5 pp, medium confidence
16Claude Opus 563.0%estimated ± 5.1 pp, medium confidence
17GPT-5.6 Cyber62.3%estimated ± 4.5 pp, medium confidence
18Qwen3.8 Max Preview61.1%estimated ± 6.4 pp, medium confidence
19Qwen3.8-Flash-Next59.9%estimated ± 6.4 pp, medium confidence
20Ling 3.1 Flash59.6%estimated ± 6.4 pp, medium confidence
21Claude Fable 559.5%estimated ± 5.1 pp, medium confidence
22DeepSeek V4 Pro58.7%estimated ± 6.4 pp, medium confidence
23Step 5 Preview58.6%estimated ± 6.4 pp, medium confidence
24Claude Fable 5.158.0%measured
25Claude Mythos 5.157.8%estimated ± 4.5 pp, medium confidence
26GPT-5.6 Terra57.7%estimated ± 4.5 pp, medium confidence
27Claude Sonnet 5.557.4%estimated ± 5.1 pp, medium confidence
28Gemini 3.8 Flash57.4%estimated ± 5.1 pp, medium confidence
29Gemini 4 Hydrogen57.4%estimated ± 5.1 pp, medium confidence
30Claude Haiku 5.557.4%estimated ± 5.1 pp, medium confidence
31Gemini 3.1 Pro57.4%estimated ± 5.1 pp, medium confidence
32Gemini 3.7 Flash57.4%estimated ± 5.1 pp, medium confidence
33Kimi K357.4%estimated ± 5.1 pp, medium confidence
34Claude Haiku 4.557.4%estimated ± 5.1 pp, medium confidence
35Claude Opus 4.857.4%estimated ± 5.1 pp, medium confidence
36Claude Sonnet 4.657.4%estimated ± 5.1 pp, medium confidence
37Claude Sonnet 557.4%estimated ± 5.1 pp, medium confidence
38DeepSeek V4 Flash 073157.4%estimated ± 5.1 pp, medium confidence
39Gemini 3.5 Flash57.4%estimated ± 5.1 pp, medium confidence
40Gemini 3.6 Flash57.4%estimated ± 5.1 pp, medium confidence
41GLM-5.257.4%estimated ± 5.1 pp, medium confidence
42GLM-5.3-Flash57.4%estimated ± 5.1 pp, medium confidence
43GPT-5.3-Codex-Spark57.4%estimated ± 5.1 pp, medium confidence
44Grok 4.2057.4%estimated ± 5.1 pp, medium confidence
45Llama 4 Maverick57.4%estimated ± 5.1 pp, medium confidence
46MiMo-V2.5-Pro57.4%estimated ± 5.1 pp, medium confidence
47MiniMax M357.4%estimated ± 5.1 pp, medium confidence
48Muse Ember57.4%estimated ± 5.1 pp, medium confidence
49Muse Spark 1.257.4%estimated ± 5.1 pp, medium confidence
50Muse Spark Open57.4%estimated ± 5.1 pp, medium confidence
51Muse Wonder57.4%estimated ± 5.1 pp, medium confidence
52Nemotron 3 Ultra57.4%estimated ± 5.1 pp, medium confidence
53Qwen3.8-27B57.4%estimated ± 5.1 pp, medium confidence
54Qwen3.8 Max57.4%estimated ± 5.1 pp, medium confidence
55Grok 4.657.0%measured
56Kimi K3.1 Pro56.4%estimated ± 4.5 pp, medium confidence
57DeepSeek V4.1 Flash 101755.0%measured
58Muse Spark 1.355.0%measured
59MiMo-V2.6-Pro55.0%estimated ± 4.5 pp, medium confidence
60Claude Mythos Preview54.8%estimated ± 4.5 pp, medium confidence
61MiniMax M3.1 Flash54.2%estimated ± 6.4 pp, medium confidence
62GLM-5.353.5%estimated ± 4.5 pp, low confidence
63Mistral Large 453.4%estimated ± 5.9 pp, low confidence
64Hy4 preview53.0%measured
65GPT-5.552.7%estimated ± 4.5 pp, low confidence
66GPT-5.6 Luna52.2%estimated ± 4.5 pp, low confidence
67GPT-6 Sol52.0%measured
68GPT-6 Luna51.7%estimated ± 4.5 pp, low confidence
69Grok 4.551.5%estimated ± 6.4 pp, medium confidence
70Claude Opus 4.750.3%estimated ± 6.4 pp, medium confidence
71GPT-5.448.9%estimated ± 4.5 pp, low confidence
72MiMo-V2.6-Flash48.9%estimated ± 4.5 pp, low confidence
73Muse Spark 1.146.2%estimated ± 4.5 pp, low confidence
74Hy3 Preview44.7%estimated ± 6.4 pp, medium confidence
75Apodex 1.143.5%estimated ± 6.4 pp, medium confidence
76Apodex 1.1 Mini43.5%estimated ± 6.4 pp, medium confidence
77Quasar 438B43.3%estimated ± 6.4 pp, medium confidence
78Beam 1.0 501B-A23B43.1%estimated ± 6.4 pp, medium confidence
79Nemotron 3.5 Super VL 120B A12B42.4%estimated ± 6.4 pp, medium confidence
80Nemotron 3.5 Super VL 120B A12B NVFP442.4%estimated ± 6.4 pp, medium confidence
81Ling 3.0 Flash VL42.3%estimated ± 6.4 pp, medium confidence
82Qwen3.7 Max40.8%estimated ± 6.4 pp, medium confidence
83Inkling-Small40.6%estimated ± 6.4 pp, medium confidence
84GLM-5.140.2%estimated ± 6.4 pp, medium confidence
85Solar Pro 439.7%estimated ± 6.4 pp, medium confidence
86Grok 4.338.7%estimated ± 6.4 pp, medium confidence
87Hy337.7%estimated ± 6.4 pp, low confidence
88Inkling37.0%measured
89Kimi K2.636.4%estimated ± 6.4 pp, low confidence
90Kimi K2.7 Code36.4%estimated ± 6.4 pp, low confidence
91Step 3.7 Flash35.3%estimated ± 6.4 pp, low confidence
92GLM-4.735.2%estimated ± 6.4 pp, low confidence
93GPT-5.4 mini35.1%estimated ± 6.4 pp, low confidence
94MiniMax M2.735.0%estimated ± 6.4 pp, low confidence
95Muse Spark34.3%estimated ± 6.4 pp, low confidence
96Qwen3.6 Plus33.9%estimated ± 6.4 pp, low confidence
97Qwen3.6-27B33.7%estimated ± 6.4 pp, low confidence
98Gemini 3.5 Flash-Lite33.6%estimated ± 6.4 pp, low confidence
99Nemotron 3.5 Bolt 74B-A9B33.4%estimated ± 6.4 pp, low confidence
100A.X K232.3%estimated ± 6.4 pp, low confidence
101GPT-5.4 nano31.6%estimated ± 6.4 pp, low confidence
102Ling 3.0 Flash31.6%estimated ± 6.4 pp, low confidence
103GPT-529.9%estimated ± 6.4 pp, low confidence
104Qwen3.6-35B-A3B28.4%estimated ± 6.4 pp, low confidence
105Kimi K2.525.3%estimated ± 6.4 pp, low confidence
106GPT-5.124.5%estimated ± 6.4 pp, low confidence
107Qwen3.5-122B-A10B23.7%estimated ± 6.4 pp, low confidence
108K-EXAONE 2.023.3%estimated ± 6.4 pp, low confidence
109Muse Glimmer 30B21.8%estimated ± 6.4 pp, low confidence
110Qwen3.7 Plus20.6%estimated ± 6.4 pp, low confidence
111Mistral Medium 3.5 128B20.2%estimated ± 6.4 pp, low confidence
112MiniCPM5-2B16.6%estimated ± 6.4 pp, low confidence
113Nemotron 3.5 Lightning 30B A3B NVFP411.5%estimated ± 6.4 pp, low confidence
114MiMo-V2-Flash10.7%estimated ± 6.4 pp, low confidence
115Gemma 4 31B9.9%estimated ± 6.4 pp, low confidence
116GPT-OSS 120B9.1%estimated ± 6.4 pp, low confidence
117Granite 4.2 30B6.4%estimated ± 6.4 pp, low confidence
118Gemma 4 26B A4B5.6%estimated ± 6.4 pp, low confidence
119Ling 3.0 Tiny5.2%estimated ± 6.4 pp, low confidence
120Command A+1.3%estimated ± 6.4 pp, low confidence
121Celeris-10.0%estimated ± 6.4 pp, low confidence
122Claude 3.5 Haiku0.0%estimated ± 6.4 pp, low confidence
123DeepSeek V30.0%estimated ± 6.4 pp, low confidence
124DeepSeek V3 03240.0%estimated ± 6.4 pp, low confidence
125Gemini 2.5 Pro0.0%estimated ± 6.4 pp, low confidence
126Gemma 3 27B0.0%estimated ± 6.4 pp, low confidence
127Gemma 4 12B0.0%estimated ± 6.4 pp, low confidence
128Gemma 4 E2B0.0%estimated ± 6.4 pp, low confidence
129Gemma 4 E4B0.0%estimated ± 6.4 pp, low confidence
130GPT-4.1 mini0.0%estimated ± 6.4 pp, low confidence
131GPT-4.1 nano0.0%estimated ± 6.4 pp, low confidence
132GPT-4o0.0%estimated ± 6.4 pp, low confidence
133GPT-4o mini0.0%estimated ± 6.4 pp, low confidence
134GPT-OSS 20B0.0%estimated ± 6.4 pp, low confidence
135Granite 4.2 3B0.0%estimated ± 6.4 pp, low confidence
136Granite 4.2 8B0.0%estimated ± 6.4 pp, low confidence
137K-Exaone0.0%estimated ± 6.4 pp, low confidence
138LFM2.5-2.6B0.0%estimated ± 6.4 pp, low confidence
139Ling 2.6 Flash0.0%estimated ± 6.4 pp, low confidence
140Llama 3.3 70B0.0%estimated ± 6.4 pp, low confidence
141Llama 4 Scout0.0%estimated ± 6.4 pp, low confidence
142Mercury 2.50.0%estimated ± 6.4 pp, low confidence
143Mistral Large 30.0%estimated ± 6.4 pp, low confidence
144Mistral Small 3.10.0%estimated ± 6.4 pp, low confidence
145Mistral Small 40.0%estimated ± 6.4 pp, low confidence
146Nemotron 3 Nano 30B0.0%estimated ± 6.4 pp, low confidence
147Nemotron 3 Nano Omni 30B A3B0.0%estimated ± 6.4 pp, low confidence
148North Mini Code0.0%estimated ± 6.4 pp, low confidence
149Solar Pro 30.0%estimated ± 6.4 pp, low confidence
150Trinity-Large-Preview0.0%estimated ± 6.4 pp, low confidence
151Trinity-Large-Thinking0.0%estimated ± 6.4 pp, low confidence
152Ultravox v0.6 Llama 3.3 70B0.0%estimated ± 6.4 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General