benchgap
General

StealthMark leaderboard

As of 2026-11-18, the highest measured score on StealthMark is 74.6% by GPT-6 Astra. 91 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Gemini 4 Carbon74.8%estimated ± 2.8 pp, medium confidence
2Claude Fable 5.274.7%estimated ± 6.5 pp, low confidence
3GPT-6 Astra74.6%measured
4Claude Opus 5.572.6%measured
5Claude Sonnet 5.570.2%measured
6GPT-6.2 Astra68.6%estimated ± 6.1 pp, medium confidence
7GPT-6.2 Sol68.6%estimated ± 6.1 pp, medium confidence
8GPT-6 Cyber68.6%estimated ± 6.1 pp, medium confidence
9GPT-6.1 Sol67.8%measured
10Gemini 4 Argon67.4%measured
11Claude Mythos 5.265.5%estimated ± 6.1 pp, medium confidence
12Claude Opus 564.6%estimated ± 3.5 pp, high confidence
13Claude Fable 5.162.9%measured
14Qwen3.8 Max62.4%measured
15GPT-5.5 Pro60.6%estimated ± 9.7 pp, low confidence
16GPT-6.1 Terra60.4%measured
17DeepSeek V4.1 Flash 101759.1%measured
18GPT-5.4 Pro59.0%estimated ± 9.7 pp, low confidence
19Gemini 3 Pro58.3%estimated ± 10.7 pp, low confidence
20Claude Opus 4.6 (Adaptive)58.3%estimated ± 10.7 pp, low confidence
21Grok 4.20 Multi-agent58.3%estimated ± 10.7 pp, low confidence
22Grok 4.1 Fast (Reasoning)58.3%estimated ± 10.7 pp, low confidence
23Hy458.3%estimated ± 10.7 pp, low confidence
24Muse Wonder57.0%measured
25Kimi K3.1 Pro56.7%measured
26Claude Mythos 556.5%estimated ± 6.1 pp, medium confidence
27GPT-5.6 Sol53.7%measured
28GLM-5.352.9%measured
29Claude Mythos Preview52.8%estimated ± 6.1 pp, medium confidence
30GPT-6 Sol52.0%measured
31Qwen3.8-27B51.5%measured
32Kimi K350.6%measured
33Claude Haiku 5.549.5%measured
34MiMo-V2.6-Flash49.3%measured
35Grok 4.648.9%estimated ± 6.0 pp, medium confidence
36GLM-5.3-Flash47.9%estimated ± 7.7 pp, medium confidence
37Gemini 3.8 Flash47.9%measured
38MiMo-V2.6-Pro47.9%measured
39GLM-5.247.6%estimated ± 8.7 pp, medium confidence
40Grok 4.747.2%measured
41SWE-2 Flash46.5%estimated ± 6.1 pp, medium confidence
42GPT-5.246.1%estimated ± 6.5 pp, medium confidence
43GPT-5.446.0%estimated ± 6.5 pp, medium confidence
44GPT-5.6 Terra45.8%measured
45Claude Fable 545.8%estimated ± 6.5 pp, medium confidence
46Claude Opus 4.745.6%estimated ± 6.5 pp, medium confidence
47MiniMax M3.1 Flash45.4%measured
48Claude Opus 4.545.3%estimated ± 6.5 pp, medium confidence
49Claude Opus 4.645.3%estimated ± 6.5 pp, medium confidence
50Gemini 3.7 Flash45.3%estimated ± 6.5 pp, medium confidence
51Gemini 3.6 Flash45.2%estimated ± 6.5 pp, medium confidence
52Grok 4.545.2%estimated ± 6.5 pp, medium confidence
53Kimi K2.645.2%estimated ± 6.5 pp, medium confidence
54Gemini 3.5 Flash45.2%estimated ± 9.2 pp, medium confidence
55Qwen3.7 Max45.2%estimated ± 9.2 pp, medium confidence
56Muse Spark 1.144.9%estimated ± 9.7 pp, low confidence
57Muse Spark Open44.1%measured
58Hy4 preview43.9%measured
59o343.6%estimated ± 9.2 pp, medium confidence
60Mistral Large 443.3%measured
61DeepSeek V4 Pro42.8%measured
62Claude Opus 4.842.4%estimated ± 3.5 pp, high confidence
63Ling 3.1 Flash42.4%measured
64Grok 4.2042.1%estimated ± 9.7 pp, low confidence
65Claude Opus 4.5 Thinking41.3%estimated ± 10.7 pp, low confidence
66Claude Sonnet 4.5 Thinking41.3%estimated ± 10.7 pp, low confidence
67DeepSeek V3.241.3%estimated ± 10.7 pp, low confidence
68Gemini 3.1 Flash-Lite41.3%estimated ± 10.7 pp, low confidence
69Grok 4.1 Fast41.3%estimated ± 10.7 pp, low confidence
70Hy341.3%estimated ± 10.7 pp, low confidence
71Hy3 Preview41.3%estimated ± 10.7 pp, low confidence
72Llama 4 Maverick41.3%estimated ± 10.7 pp, low confidence
73MiMo-V2.5-Pro41.3%estimated ± 10.7 pp, low confidence
74MiMo-V2-Pro41.3%estimated ± 10.7 pp, low confidence
75MiniMax M2.741.3%estimated ± 10.7 pp, low confidence
76Mistral Large 341.3%estimated ± 10.7 pp, low confidence
77Mistral Medium 3.5 128B41.3%estimated ± 10.7 pp, low confidence
78Qwen3.5-27B41.3%estimated ± 10.7 pp, low confidence
79Step 3.7 Flash41.3%estimated ± 10.7 pp, low confidence
80Trinity-Large-Thinking41.3%estimated ± 10.7 pp, low confidence
81Claude Sonnet 541.0%estimated ± 3.5 pp, high confidence
82Seed 2.1 Turbo40.4%estimated ± 9.7 pp, low confidence
83Gemini 3.1 Pro40.1%estimated ± 2.8 pp, medium confidence
84Muse Spark 1.339.5%measured
85GPT-6 Luna39.1%measured
86Step 5 Preview39.0%measured
87Qwen3.5 Flash38.9%estimated ± 9.2 pp, medium confidence
88MiniMax M338.6%measured
89Qwen 3.6 Max (preview)38.4%estimated ± 9.2 pp, medium confidence
90Qwen3.5 397B37.8%estimated ± 9.2 pp, medium confidence
91GPT-5.6 Luna37.8%measured
92GLM-4.737.5%estimated ± 9.7 pp, low confidence
93Inkling37.5%estimated ± 9.7 pp, low confidence
94Seed 2.1 Pro37.5%estimated ± 9.7 pp, low confidence
95Claude Sonnet 4.537.3%estimated ± 9.2 pp, medium confidence
96Qwen3.5-122B-A10B37.3%estimated ± 9.2 pp, medium confidence
97Qwen3.5 Plus37.3%estimated ± 9.2 pp, medium confidence
98Qwen3.7 Plus37.3%estimated ± 9.2 pp, medium confidence
99GLM-5.137.0%estimated ± 9.7 pp, low confidence
100Claude Sonnet 4.636.8%estimated ± 9.2 pp, medium confidence
101GPT-5.136.8%estimated ± 9.2 pp, medium confidence
102GPT-5.536.5%measured
103Qwen3.7 Flash36.3%estimated ± 9.2 pp, medium confidence
104Grok 4.335.9%estimated ± 9.7 pp, low confidence
105GPT-535.8%estimated ± 8.7 pp, low confidence
106GPT-4o mini34.7%estimated ± 9.2 pp, medium confidence
107Qwen3.6 Plus34.7%estimated ± 9.2 pp, medium confidence
108Kimi K2.7 Code33.6%estimated ± 9.7 pp, low confidence
109Qwen3.6-27B32.4%measured
110GPT-4.1 mini32.1%estimated ± 9.2 pp, medium confidence
111GPT-5.4 mini32.1%estimated ± 9.2 pp, medium confidence
112o3-mini32.1%estimated ± 9.2 pp, medium confidence
113GPT-4.131.9%estimated ± 9.7 pp, low confidence
114Beam 1.0 501B-A23B31.9%measured
115o131.4%estimated ± 9.7 pp, low confidence
116GPT-5.4 nano31.0%estimated ± 9.2 pp, low confidence
117o4-mini31.0%estimated ± 9.2 pp, low confidence
118Qwen3 Max31.0%estimated ± 9.2 pp, low confidence
119GPT-4o30.8%estimated ± 9.7 pp, low confidence
120Gemini 3.5 Flash-Lite30.5%estimated ± 3.5 pp, medium confidence
121Gemini 3 Flash30.5%estimated ± 3.5 pp, medium confidence
122Muse Spark 1.230.3%estimated ± 7.7 pp, low confidence
123Claude Haiku 4.530.2%estimated ± 9.7 pp, low confidence
124Gemma 4 31B29.5%measured
125GPT-OSS 120B28.4%estimated ± 9.2 pp, low confidence
126Claude 3 Opus28.0%estimated ± 9.7 pp, low confidence
127Inkling-Small27.5%measured
128GPT-4.1 nano25.7%estimated ± 9.7 pp, low confidence
129Gemma 4 26B A4B21.2%measured
130Qwen3.6-35B-A3B20.8%measured
131DeepSeek V4 Flash 073120.6%estimated ± 7.7 pp, low confidence
132Nemotron 3 Ultra16.7%measured
133MAI-Thinking-114.1%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General