Agentic · tools

AA Cyber Index leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

As of 2026-10-29, the highest measured score on AA Cyber Index is 64.1% by Kimi K3.1 Pro. 114 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Kimi K3.1 Pro	64.1%	measured
2	Grok 4.7	56.4%	measured
3	MiMo-V2.6-Pro	56.1%	measured
4	GPT-6 Luna	52.7%	measured
5	GLM-5.3-Flash	49.9%	measured
6	Mistral Large 4	49.5%	measured
7	Step 5 Preview	47.7%	measured
8	MiniMax M3.1 Flash	46.3%	estimated ± 9.4 pp, low confidence
9	Gemini 4 Carbon	46.3%	estimated ± 9.4 pp, medium confidence
10	Atria Dawn Preview	46.3%	estimated ± 9.4 pp, medium confidence
11	Ling 3.1 Flash	46.3%	estimated ± 9.4 pp, medium confidence
12	MiMo-V2.6-Flash	46.3%	estimated ± 9.4 pp, medium confidence
13	Gemini 4 Argon	46.3%	estimated ± 9.4 pp, medium confidence
14	Ultramarine 1 Flash	46.3%	estimated ± 9.4 pp, medium confidence
15	GPT-5.5 Pro	42.2%	estimated ± 13.3 pp, low confidence
16	GPT-5.4 Pro	41.9%	estimated ± 13.3 pp, low confidence
17	Claude Sonnet 5.5	41.5%	estimated ± 12.8 pp, low confidence
18	Claude Mythos 5	41.4%	estimated ± 13.3 pp, low confidence
19	Kimi K3	41.3%	measured
20	DeepSeek V4.1 Flash 1017	40.8%	measured
21	Claude Opus 4.6	39.8%	estimated ± 13.3 pp, low confidence
22	Muse Spark Open	39.7%	estimated ± 12.8 pp, low confidence
23	GPT-5.4	39.4%	estimated ± 13.3 pp, low confidence
24	Claude Haiku 5.5	37.8%	estimated ± 12.8 pp, low confidence
25	Ling 3.0 Flash Sante	37.6%	estimated ± 12.8 pp, low confidence
26	Nemotron 3.5 Super VL 120B A12B	37.6%	estimated ± 13.3 pp, low confidence
27	Nemotron 3.5 Super VL 120B A12B NVFP4	37.4%	estimated ± 13.3 pp, low confidence
28	Apodex 1.1	37.1%	estimated ± 12.8 pp, low confidence
29	Ornith-1.5-397B	37.1%	estimated ± 12.8 pp, low confidence
30	GPT-6 Sol	36.8%	measured
31	Sage-A1.5	36.5%	estimated ± 12.8 pp, low confidence
32	GLM-5.3	36.4%	measured
33	dots3-note Preview	35.2%	estimated ± 12.8 pp, low confidence
34	Nemotron 3.5 Bolt 74B-A9B BF16	34.4%	estimated ± 13.3 pp, low confidence
35	Beam 1.0 501B-A23B FP8	34.3%	estimated ± 12.8 pp, low confidence
36	Beam 1.0 501B-A23B NVFP4	34.3%	estimated ± 12.8 pp, low confidence
37	Qwen3.5 397B	34.0%	estimated ± 13.3 pp, low confidence
38	GLM-5	33.9%	estimated ± 12.8 pp, low confidence
39	LFM3-1.7B	33.9%	estimated ± 9.4 pp, medium confidence
40	Beam 1.0 501B-A23B	33.7%	measured
41	GPT-6 Astra	33.4%	measured
42	Agents-A1-4B	33.1%	estimated ± 13.3 pp, low confidence
43	Nemotron 3.5 Bolt 74B-A9B	33.0%	estimated ± 13.3 pp, low confidence
44	GPT-6.1 Sol	32.8%	estimated ± 9.4 pp, medium confidence
45	Hy4 preview	32.8%	estimated ± 9.4 pp, medium confidence
46	LFM2.5-24B-A2B	32.8%	estimated ± 9.4 pp, low confidence
47	Qwen3.8 Max	32.8%	estimated ± 9.4 pp, low confidence
48	GPT-5.2	32.6%	estimated ± 13.3 pp, low confidence
49	Agents-A1	32.4%	estimated ± 12.8 pp, low confidence
50	Step 3.7 Flash	32.1%	estimated ± 12.8 pp, low confidence
51	Claude Opus 5	32.0%	estimated ± 9.2 pp, low confidence
52	Muse Spark 1.3	31.7%	estimated ± 9.2 pp, low confidence
53	Grok 4.6	30.7%	estimated ± 9.2 pp, low confidence
54	Qwen3.5-27B	30.6%	estimated ± 13.3 pp, low confidence
55	Qwen3.5-35B-A3B	30.6%	estimated ± 13.3 pp, low confidence
56	Kimi K2.5	30.5%	estimated ± 13.3 pp, low confidence
57	Kimi K2.5 (Reasoning)	30.5%	estimated ± 13.3 pp, low confidence
58	Claude Fable 5	29.6%	estimated ± 9.2 pp, low confidence
59	GPT-5.6 Sol	29.4%	estimated ± 9.2 pp, low confidence
60	Qwen3.8 Max Preview	29.0%	estimated ± 9.2 pp, low confidence
61	DeepSeek V4 Pro 0813	29.0%	estimated ± 9.2 pp, low confidence
62	Claude Opus 5.5	28.7%	measured
63	GLM-4.7	26.7%	estimated ± 13.3 pp, low confidence
64	Claude Sonnet 5	26.5%	estimated ± 9.2 pp, low confidence
65	Muse Spark 1.2	26.4%	estimated ± 9.2 pp, low confidence
66	GPT-5.6 Terra	26.3%	estimated ± 9.2 pp, low confidence
67	Spark-X2.6-4B	26.0%	estimated ± 12.8 pp, low confidence
68	GPT-5.6 Luna	25.8%	estimated ± 9.2 pp, low confidence
69	Claude Opus 4.8	25.7%	estimated ± 9.2 pp, low confidence
70	Sarvam 105B	25.6%	estimated ± 13.3 pp, low confidence
71	Grok 4.5	25.5%	estimated ± 9.2 pp, low confidence
72	Solar Pro 4	25.5%	estimated ± 13.3 pp, low confidence
73	DeepSeek V4 Flash 0731	25.3%	estimated ± 9.2 pp, low confidence
74	LongCat-Flash-Lite-Sparse	25.2%	estimated ± 13.3 pp, low confidence
75	Gemini 3.8 Flash	25.1%	measured
76	Claude Fable 5.1	25.0%	measured
77	Claude Opus 4.7 (Adaptive)	24.2%	estimated ± 9.2 pp, low confidence
78	GLM-5.2	24.2%	estimated ± 9.2 pp, low confidence
79	Ornith-1.5-35B-A3B	23.9%	estimated ± 12.8 pp, low confidence
80	GPT-5.5	23.1%	estimated ± 9.2 pp, low confidence
81	Gemini 3.7 Flash	22.6%	estimated ± 9.2 pp, low confidence
82	MiniMax M3	22.4%	measured
83	Ornith-1.5-9B	22.0%	estimated ± 12.8 pp, low confidence
84	Quasar 438B	20.7%	estimated ± 9.2 pp, low confidence
85	Gemini 3.6 Flash	19.3%	estimated ± 9.2 pp, low confidence
86	Sarvam 30B	19.0%	estimated ± 13.3 pp, low confidence
87	Muse Spark 1.1	17.9%	estimated ± 9.2 pp, low confidence
88	Gemini 3.5 Flash	17.7%	estimated ± 9.2 pp, low confidence
89	Hy3	16.8%	estimated ± 9.2 pp, low confidence
90	Hy3 Preview	16.8%	estimated ± 9.2 pp, low confidence
91	GLM-5.1	16.6%	estimated ± 9.2 pp, low confidence
92	Inkling-Small	16.4%	estimated ± 9.2 pp, low confidence
93	Inkling	16.0%	estimated ± 9.2 pp, low confidence
94	Qwen3.7 Max	15.8%	estimated ± 9.2 pp, low confidence
95	MiMo-V2.5-Pro	15.1%	estimated ± 9.2 pp, low confidence
96	Kimi K2.7 Code	15.0%	estimated ± 9.2 pp, low confidence
97	Kimi K2.6	14.8%	estimated ± 9.2 pp, low confidence
98	Ling 3.0 Flash	14.1%	estimated ± 9.2 pp, low confidence
99	Qwen3.6-27B	13.5%	estimated ± 9.2 pp, low confidence
100	Qwen3.7 Plus	13.3%	estimated ± 9.2 pp, low confidence
101	GPT-5.4 mini	13.3%	estimated ± 9.2 pp, low confidence
102	Nemotron 3 Ultra	13.1%	measured
103	Qwen3.8-27B	12.7%	measured
104	GPT-5.4 nano	12.1%	estimated ± 9.2 pp, low confidence
105	Grok 4.3	11.8%	estimated ± 9.2 pp, low confidence
106	MiniMax M2.7	11.5%	estimated ± 9.2 pp, low confidence
107	Gemini 3.5 Flash-Lite	11.0%	estimated ± 9.2 pp, low confidence
108	Qwen3.6-35B-A3B	10.4%	estimated ± 9.2 pp, low confidence
109	Qwen3.8-Flash-Next	7.8%	estimated ± 13.3 pp, low confidence
110	Muse Glimmer 30B	7.4%	estimated ± 9.2 pp, low confidence
111	Gemini 3.1 Pro	7.3%	estimated ± 9.2 pp, low confidence
112	Qwen3.5-122B-A10B	6.9%	estimated ± 9.2 pp, low confidence
113	Mistral Medium 3.5 128B	6.7%	estimated ± 9.2 pp, low confidence
114	Claude Opus 4.7	5.3%	estimated ± 13.3 pp, low confidence
115	Gemma 4 31B	4.9%	estimated ± 9.2 pp, low confidence
116	GPT-OSS 120B	4.5%	estimated ± 9.2 pp, low confidence
117	Nemotron 3.5 Lightning 30B A3B NVFP4	4.5%	estimated ± 9.2 pp, low confidence
118	Claude Sonnet 4.6	3.1%	estimated ± 13.3 pp, low confidence
119	Granite 4.2 8B	2.7%	estimated ± 9.2 pp, low confidence
120	Command A+	2.7%	estimated ± 9.2 pp, low confidence
121	Gemini 2.5 Pro	2.6%	estimated ± 9.2 pp, low confidence
122	Mistral Large 3	1.8%	estimated ± 9.2 pp, low confidence
123	Mistral Small 4	1.1%	estimated ± 9.2 pp, low confidence
124	Mistral Small 4 (Reasoning)	1.1%	estimated ± 9.2 pp, low confidence
125	GPT-OSS 20B	1.0%	estimated ± 9.2 pp, low confidence
126	Trinity-Large-Preview	0.9%	estimated ± 9.2 pp, low confidence
127	Trinity-Large-Thinking	0.9%	estimated ± 9.2 pp, low confidence
128	Nemotron 3 Nano 30B	0.7%	estimated ± 9.2 pp, low confidence
129	DeepSeek V3	0.6%	estimated ± 9.2 pp, low confidence
130	Celeris-1	0.5%	estimated ± 9.2 pp, low confidence
131	Llama 4 Maverick	0.5%	estimated ± 9.2 pp, low confidence
132	Llama 4 Scout	0.4%	estimated ± 9.2 pp, low confidence
133	Gemma 3 27B	0.1%	estimated ± 9.2 pp, low confidence

AA Cyber Index leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General