benchgap
General

BinaryAudit leaderboard

As of 2026-11-18, the highest measured score on BinaryAudit is 49.1% by Claude Opus 4.6. 88 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-5.555.0%estimated ± 12.0 pp, low confidence
2Claude Opus 4.649.1%measured
3Gemini 3.1 Pro49.1%measured
4Claude Opus 4.747.2%estimated ± 12.0 pp, low confidence
5GPT-5.2-Codex45.6%measured
6Gemini 3 Pro43.9%measured
7GLM-5.143.8%estimated ± 12.0 pp, low confidence
8GPT-6 Astra42.8%estimated ± 8.9 pp, low confidence
9GPT-5.3 Codex42.1%measured
10Hy340.8%estimated ± 12.0 pp, low confidence
11Gemini 3.8 Flash40.5%estimated ± 8.9 pp, low confidence
12Claude Opus 4.838.8%estimated ± 12.0 pp, low confidence
13Kimi K2.638.7%estimated ± 12.0 pp, low confidence
14MiniMax M337.6%estimated ± 12.0 pp, low confidence
15Claude Opus 4.536.8%measured
16Gemini 3 Flash36.8%measured
17Claude Fable 535.8%estimated ± 12.1 pp, low confidence
18Claude Fable 5.135.8%estimated ± 12.1 pp, low confidence
19Claude Fable 5.235.8%estimated ± 12.1 pp, low confidence
20Claude Haiku 5.535.8%estimated ± 12.1 pp, low confidence
21Claude Opus 4.6 (Adaptive)35.8%estimated ± 12.1 pp, low confidence
22Claude Opus 535.8%estimated ± 12.1 pp, low confidence
23Claude Opus 5.535.8%estimated ± 12.1 pp, low confidence
24Claude Sonnet 535.8%estimated ± 12.1 pp, low confidence
25DeepSeek V4.1 Flash 101735.8%estimated ± 12.1 pp, low confidence
26Gemini 3.5 Flash-Lite35.8%estimated ± 12.1 pp, low confidence
27Gemini 3.6 Flash35.8%estimated ± 12.1 pp, low confidence
28Gemini 3.7 Flash35.8%estimated ± 12.1 pp, low confidence
29Gemini 4 Argon35.8%estimated ± 12.1 pp, low confidence
30Gemini 4 Carbon35.8%estimated ± 12.1 pp, low confidence
31Gemma 4 31B35.8%estimated ± 12.1 pp, low confidence
32GLM-5.235.8%estimated ± 12.1 pp, low confidence
33GLM-5.335.8%estimated ± 12.1 pp, low confidence
34GPT-5.6 Luna35.8%estimated ± 12.1 pp, low confidence
35GPT-5.6 Sol35.8%estimated ± 12.1 pp, low confidence
36GPT-5.6 Terra35.8%estimated ± 12.1 pp, low confidence
37GPT-6.1 Sol35.8%estimated ± 12.1 pp, low confidence
38GPT-6.1 Terra35.8%estimated ± 12.1 pp, low confidence
39GPT-6 Luna35.8%estimated ± 12.1 pp, low confidence
40GPT-6 Sol35.8%estimated ± 12.1 pp, low confidence
41Grok 4.2035.8%estimated ± 12.1 pp, low confidence
42Grok 4.20 Multi-agent35.8%estimated ± 12.1 pp, low confidence
43Grok 4.535.8%estimated ± 12.1 pp, low confidence
44Grok 4.635.8%estimated ± 12.1 pp, low confidence
45Grok 4.735.8%estimated ± 12.1 pp, low confidence
46Hy435.8%estimated ± 12.1 pp, low confidence
47Hy4 preview35.8%estimated ± 12.1 pp, low confidence
48Kimi K3.1 Pro35.8%estimated ± 12.1 pp, low confidence
49Kimi K335.8%estimated ± 12.1 pp, low confidence
50Muse Spark 1.135.8%estimated ± 12.1 pp, low confidence
51Muse Spark 1.235.8%estimated ± 12.1 pp, low confidence
52Muse Spark 1.335.8%estimated ± 12.1 pp, low confidence
53Muse Spark Open35.8%estimated ± 12.1 pp, low confidence
54Muse Wonder35.8%estimated ± 12.1 pp, low confidence
55Qwen 3.6 Max (preview)35.8%estimated ± 12.1 pp, low confidence
56Qwen3.6 Plus35.8%estimated ± 12.1 pp, low confidence
57Qwen3.7 Max35.8%estimated ± 12.1 pp, low confidence
58Qwen3.7 Plus35.8%estimated ± 12.1 pp, low confidence
59Qwen3.8-27B35.8%estimated ± 12.1 pp, low confidence
60Qwen3.8 Max35.8%estimated ± 12.1 pp, low confidence
61Step 5 Preview35.8%estimated ± 12.1 pp, low confidence
62Claude Opus 4.5 Thinking35.8%estimated ± 12.1 pp, low confidence
63Qwen3.5-122B-A10B35.8%estimated ± 12.1 pp, low confidence
64Qwen3.5-27B35.8%estimated ± 12.1 pp, low confidence
65Qwen3.7 Flash35.8%estimated ± 12.1 pp, low confidence
66Qwen3.6-35B-A3B35.8%estimated ± 12.1 pp, low confidence
67Step 3.7 Flash35.8%estimated ± 12.1 pp, low confidence
68Claude Sonnet 4.5 Thinking35.1%estimated ± 12.1 pp, low confidence
69o335.0%estimated ± 8.9 pp, low confidence
70Claude Sonnet 5.534.2%estimated ± 8.9 pp, low confidence
71GPT-5.133.7%estimated ± 8.8 pp, medium confidence
72Gemini 3.5 Flash32.4%estimated ± 12.0 pp, low confidence
73Claude Sonnet 4.631.4%estimated ± 12.0 pp, low confidence
74Gemini 2.5 Flash26.7%estimated ± 8.9 pp, low confidence
75Claude Sonnet 4.526.3%measured
76Grok 4.325.9%estimated ± 12.0 pp, low confidence
77GPT-5.4 mini25.6%estimated ± 12.0 pp, low confidence
78GLM-4.724.6%measured
79Kimi K2.524.6%measured
80o4-mini23.9%estimated ± 8.9 pp, low confidence
81Grok 4.1 Fast (Reasoning)23.2%estimated ± 8.9 pp, low confidence
82Grok 4 Fast (Reasoning)21.2%estimated ± 8.9 pp, low confidence
83Gemini 2.5 Pro21.1%measured
84MiniMax M2.719.6%estimated ± 12.0 pp, low confidence
85GLM-519.3%measured
86GPT-5.217.5%measured
87GPT-5.417.5%measured
88MiMo-V2.5-Pro12.4%estimated ± 12.1 pp, low confidence
89DeepSeek V3.212.3%measured
90GPT-5.1-Codex-Max11.4%estimated ± 8.8 pp, medium confidence
91Claude Haiku 4.510.5%measured
92MiniMax M3.1 Flash9.9%estimated ± 12.1 pp, low confidence
93Mistral Large 49.9%estimated ± 12.1 pp, low confidence
94MiMo-V2-Pro9.9%estimated ± 12.1 pp, low confidence
95Hy3 Preview9.9%estimated ± 12.1 pp, low confidence
96Trinity-Large-Thinking9.9%estimated ± 12.1 pp, low confidence
97Llama 4 Maverick9.9%estimated ± 12.1 pp, low confidence
98Mistral Large 39.9%estimated ± 12.1 pp, low confidence
99Mistral Medium 3.5 128B9.9%estimated ± 12.1 pp, low confidence
100Grok 48.8%measured
101Qwen3 235B 25078.2%estimated ± 8.9 pp, low confidence
102Gemini 3.1 Flash-Lite7.4%estimated ± 12.0 pp, low confidence
103Grok 4.1 Fast7.0%measured
104Claude 3.5 Haiku0.0%estimated ± 8.9 pp, low confidence
105Claude 3.5 Sonnet0.0%estimated ± 8.9 pp, low confidence
106Gemini 2.0 Flash0.0%estimated ± 8.9 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General