Agentic · tools

CWE-bench v1 leaderboard

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

As of 2026-10-07, the highest measured score on CWE-bench v1 is 68.0% by Gemini 4 Argon. 114 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Gemini 4 Argon	68.0%	measured
2	GPT-6 Astra	68.0%	measured
3	Grok 4.7	68.0%	measured
4	Claude Opus 5.5	67.0%	measured
5	Claude Sonnet 5.5	65.0%	estimated ± 6.0 pp, medium confidence
6	Claude Opus 5	62.7%	estimated ± 6.9 pp, medium confidence
7	Qwen3.8 Max Preview	61.2%	estimated ± 6.9 pp, medium confidence
8	GPT-6.1 Sol	60.2%	estimated ± 6.0 pp, medium confidence
9	Qwen3.8-Flash-Next	60.0%	estimated ± 6.9 pp, medium confidence
10	Ling 3.1 Flash	59.7%	estimated ± 6.9 pp, medium confidence
11	Claude Fable 5	59.4%	estimated ± 6.9 pp, medium confidence
12	GPT-5.6 Sol	59.4%	estimated ± 6.9 pp, medium confidence
13	MiMo-V2.6-Flash	59.4%	estimated ± 6.9 pp, medium confidence
14	DeepSeek V4 Pro 0813	58.8%	estimated ± 6.9 pp, medium confidence
15	GLM-5.3	58.1%	estimated ± 6.0 pp, medium confidence
16	Claude Fable 5.1	58.0%	measured
17	Grok 4.6	57.0%	measured
18	GLM-5.3-Flash	56.6%	estimated ± 6.0 pp, medium confidence
19	Gemini 3.8 Flash	56.2%	estimated ± 6.0 pp, medium confidence
20	Mistral Large 4	56.2%	estimated ± 6.0 pp, medium confidence
21	MiMo-V2.6-Pro	55.1%	estimated ± 6.0 pp, medium confidence
22	Muse Spark 1.2	55.1%	estimated ± 6.9 pp, medium confidence
23	DeepSeek V4.1 Flash	55.0%	measured
24	Muse Spark 1.3	55.0%	measured
25	Kimi K3	54.9%	estimated ± 6.0 pp, medium confidence
26	Claude Sonnet 5	54.7%	estimated ± 6.9 pp, medium confidence
27	GPT-5.6 Luna	54.6%	estimated ± 6.9 pp, medium confidence
28	Claude Opus 4.8	54.3%	estimated ± 6.9 pp, medium confidence
29	GPT-5.6 Terra	54.3%	estimated ± 6.9 pp, medium confidence
30	DeepSeek V4 Flash 0731	53.7%	estimated ± 6.9 pp, medium confidence
31	Hy4 preview	53.0%	measured
32	Gemini 3.7 Flash	52.1%	estimated ± 6.9 pp, medium confidence
33	GPT-6 Sol	52.0%	measured
34	Grok 4.5	52.0%	estimated ± 6.9 pp, medium confidence
35	GLM-5.2	51.4%	estimated ± 6.9 pp, medium confidence
36	Claude Opus 4.7 (Adaptive)	50.7%	estimated ± 6.9 pp, medium confidence
37	GPT-5.5	50.7%	estimated ± 6.9 pp, medium confidence
38	GPT-6 Luna	50.5%	estimated ± 6.0 pp, medium confidence
39	Gemini 3.5 Flash	50.3%	estimated ± 6.9 pp, medium confidence
40	Step 5 Preview	48.7%	estimated ± 6.0 pp, medium confidence
41	Gemini 3.6 Flash	48.0%	estimated ± 6.9 pp, medium confidence
42	Qwen3.8-27B	46.6%	estimated ± 6.0 pp, medium confidence
43	GPT-5.4	46.5%	estimated ± 6.9 pp, medium confidence
44	Hy3 Preview	45.1%	estimated ± 6.9 pp, medium confidence
45	Muse Spark 1.1	45.0%	estimated ± 6.9 pp, medium confidence
46	Apodex 1.1	44.3%	estimated ± 6.9 pp, medium confidence
47	Apodex 1.1 Mini	44.3%	estimated ± 6.9 pp, medium confidence
48	Quasar 438B	43.7%	estimated ± 6.9 pp, medium confidence
49	Ling 3.0 Flash VL	42.9%	estimated ± 6.9 pp, medium confidence
50	Qwen3.7 Max	41.4%	estimated ± 6.9 pp, medium confidence
51	Inkling-Small	41.0%	estimated ± 6.9 pp, medium confidence
52	MiMo-V2.5-Pro	41.0%	estimated ± 6.9 pp, medium confidence
53	GLM-5.1	40.8%	estimated ± 6.9 pp, medium confidence
54	Solar Pro 4	40.4%	estimated ± 6.9 pp, medium confidence
55	Claude Haiku 5.5	39.6%	estimated ± 6.0 pp, medium confidence
56	Grok 4.3	39.1%	estimated ± 6.9 pp, medium confidence
57	Hy3	38.3%	estimated ± 6.9 pp, low confidence
58	MiniMax M3	37.2%	estimated ± 6.0 pp, medium confidence
59	Muse Glimmer 30B	37.0%	estimated ± 6.0 pp, medium confidence
60	Nemotron 3 Ultra	37.0%	estimated ± 6.0 pp, low confidence
61	Inkling	37.0%	measured
62	Kimi K2.6	37.0%	estimated ± 6.9 pp, low confidence
63	Kimi K2.7 Code	37.0%	estimated ± 6.9 pp, low confidence
64	GLM-4.7	35.7%	estimated ± 6.9 pp, low confidence
65	GPT-5.4 mini	35.7%	estimated ± 6.9 pp, low confidence
66	Step 3.7 Flash	35.7%	estimated ± 6.9 pp, low confidence
67	MiniMax M2.7	35.6%	estimated ± 6.9 pp, low confidence
68	Muse Spark	35.0%	estimated ± 6.9 pp, low confidence
69	Qwen3.6 Plus	34.6%	estimated ± 6.9 pp, low confidence
70	Qwen3.6-27B	34.3%	estimated ± 6.9 pp, low confidence
71	Gemini 3.5 Flash-Lite	34.2%	estimated ± 6.9 pp, low confidence
72	A.X K2	33.0%	estimated ± 6.9 pp, low confidence
73	GPT-5.4 nano	32.2%	estimated ± 6.9 pp, low confidence
74	Ling 3.0 Flash	32.1%	estimated ± 6.9 pp, low confidence
75	Ling 3.0 Flash FP8	32.1%	estimated ± 6.9 pp, low confidence
76	GPT-5 (high)	30.6%	estimated ± 6.9 pp, low confidence
77	Qwen3.6-35B-A3B	29.2%	estimated ± 6.9 pp, low confidence
78	Kimi K2.5	26.1%	estimated ± 6.9 pp, low confidence
79	Kimi K2.5 (Reasoning)	26.1%	estimated ± 6.9 pp, low confidence
80	GPT-5.1	25.2%	estimated ± 6.9 pp, low confidence
81	Qwen3.5-122B-A10B	24.3%	estimated ± 6.9 pp, low confidence
82	Gemini 3.1 Pro	22.9%	estimated ± 6.9 pp, low confidence
83	Qwen3.7 Plus	21.3%	estimated ± 6.9 pp, low confidence
84	Mistral Medium 3.5 128B	20.9%	estimated ± 6.9 pp, low confidence
85	MiniCPM5-2B	17.2%	estimated ± 6.9 pp, low confidence
86	Nemotron 3.5 Lightning 30B A3B NVFP4	11.9%	estimated ± 6.9 pp, low confidence
87	MiMo-V2-Flash	11.3%	estimated ± 6.9 pp, low confidence
88	Gemma 4 31B	10.5%	estimated ± 6.9 pp, low confidence
89	GPT-OSS 120B	9.7%	estimated ± 6.9 pp, low confidence
90	Granite 4.2 30B	6.9%	estimated ± 6.9 pp, low confidence
91	Gemma 4 26B A4B	6.1%	estimated ± 6.9 pp, low confidence
92	Ling 3.0 Tiny	5.7%	estimated ± 6.9 pp, low confidence
93	Command A+	1.7%	estimated ± 6.9 pp, low confidence
94	Celeris-1	0.0%	estimated ± 6.9 pp, low confidence
95	DeepSeek V3	0.0%	estimated ± 6.9 pp, low confidence
96	DeepSeek V3 0324	0.0%	estimated ± 6.9 pp, low confidence
97	Gemini 2.5 Pro	0.0%	estimated ± 6.9 pp, low confidence
98	Gemma 3 27B	0.0%	estimated ± 6.9 pp, low confidence
99	Gemma 4 12B	0.0%	estimated ± 6.9 pp, low confidence
100	Gemma 4 E2B	0.0%	estimated ± 6.9 pp, low confidence
101	Gemma 4 E4B	0.0%	estimated ± 6.9 pp, low confidence
102	GPT-4.1 mini	0.0%	estimated ± 6.9 pp, low confidence
103	GPT-4.1 nano	0.0%	estimated ± 6.9 pp, low confidence
104	GPT-4o	0.0%	estimated ± 6.9 pp, low confidence
105	GPT-4o mini	0.0%	estimated ± 6.9 pp, low confidence
106	GPT-OSS 20B	0.0%	estimated ± 6.9 pp, low confidence
107	Granite 4.2 3B	0.0%	estimated ± 6.9 pp, low confidence
108	Granite 4.2 8B	0.0%	estimated ± 6.9 pp, low confidence
109	K-Exaone	0.0%	estimated ± 6.9 pp, low confidence
110	LFM2.5-2.6B	0.0%	estimated ± 6.9 pp, low confidence
111	Ling 2.6 Flash	0.0%	estimated ± 6.9 pp, low confidence
112	Llama 4 Maverick	0.0%	estimated ± 6.9 pp, low confidence
113	Llama 4 Scout	0.0%	estimated ± 6.9 pp, low confidence
114	Mercury 2.5	0.0%	estimated ± 6.9 pp, low confidence
115	Mistral Large 3	0.0%	estimated ± 6.9 pp, low confidence
116	Mistral Small 4	0.0%	estimated ± 6.9 pp, low confidence
117	Mistral Small 4 (Reasoning)	0.0%	estimated ± 6.9 pp, low confidence
118	Nemotron 3 Nano 30B	0.0%	estimated ± 6.9 pp, low confidence
119	Nemotron 3 Nano Omni 30B A3B	0.0%	estimated ± 6.9 pp, low confidence
120	Nemotron 3 Super 100B	0.0%	estimated ± 6.9 pp, low confidence
121	North Mini Code	0.0%	estimated ± 6.9 pp, low confidence
122	Solar Pro 3	0.0%	estimated ± 6.9 pp, low confidence
123	Trinity-Large-Preview	0.0%	estimated ± 6.9 pp, low confidence
124	Trinity-Large-Thinking	0.0%	estimated ± 6.9 pp, low confidence
125	Ultravox v0.6 Llama 3.3 70B	0.0%	estimated ± 6.9 pp, low confidence

CWE-bench v1 leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General