benchgap
Agentic · tools

AA Tau3 Banking leaderboard

As of 2026-11-21, the highest measured score on AA Tau3 Banking is 50.7% by Grok 4.6. 161 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Atria Dawn Preview100.0%estimated ± 3.3 pp, low confidence
2Claude Opus 5.5100.0%estimated ± 2.4 pp, low confidence
3DeepSeek V4.1 Flash 1017100.0%estimated ± 2.4 pp, low confidence
4Gemini 3.8 Flash Cyber100.0%estimated ± 3.3 pp, low confidence
5GPT-6.2 Astra100.0%estimated ± 2.4 pp, low confidence
6GPT-6.2 Sol100.0%estimated ± 2.4 pp, low confidence
7GPT-6 Astra100.0%estimated ± 2.4 pp, low confidence
8GPT-6 Sol100.0%estimated ± 2.4 pp, low confidence
9Grok 4.7100.0%estimated ± 2.4 pp, low confidence
10Kimi K3.1 Pro100.0%estimated ± 2.4 pp, low confidence
11Ling 3.1 Flash100.0%estimated ± 3.3 pp, low confidence
12MiMo-V2.6-Pro100.0%estimated ± 2.4 pp, low confidence
13Mistral Large 4100.0%estimated ± 2.4 pp, low confidence
14Fugu Cyber100.0%estimated ± 3.3 pp, low confidence
15Claude Opus 4.898.8%estimated ± 2.0 pp, low confidence
16GPT-6 Luna56.2%estimated ± 2.4 pp, low confidence
17ERNIE 5.5 Flash Preview52.0%estimated ± 3.3 pp, low confidence
18Hy451.0%estimated ± 3.3 pp, low confidence
19Grok 4.650.7%measured
20Gemini 4 Carbon50.6%estimated ± 3.3 pp, low confidence
21Muse Spark 1.350.5%measured
22GLM-5.350.3%measured
23MiMo-V2.6-Flash50.2%estimated ± 3.3 pp, low confidence
24Hy4 flash50.2%estimated ± 3.3 pp, low confidence
25LongCat-2.550.1%estimated ± 3.3 pp, low confidence
26Claude Fable 5.250.0%estimated ± 3.3 pp, low confidence
27Gemini 4 Argon50.0%estimated ± 3.3 pp, low confidence
28Ultramarine 1 Flash49.8%estimated ± 3.3 pp, low confidence
29Muse Wonder Full49.0%estimated ± 4.1 pp, medium confidence
30Claude Sonnet 5.548.8%estimated ± 4.1 pp, medium confidence
31GPT-6.1 Terra48.2%estimated ± 4.6 pp, low confidence
32Qwen3.8-27B48.0%measured
33Muse Wonder47.8%estimated ± 4.1 pp, medium confidence
34LFM3-1.7B47.4%estimated ± 3.3 pp, medium confidence
35Claude Fable 5.147.2%measured
36GLM-5.3-Flash47.2%measured
37Gemini 3.8 Flash47.1%estimated ± 2.4 pp, low confidence
38Step 5 Preview47.1%estimated ± 2.4 pp, low confidence
39Beam 1.0 501B-A23B47.1%estimated ± 2.4 pp, low confidence
40MiniMax M347.1%estimated ± 2.4 pp, low confidence
41Nemotron 3 Ultra47.1%estimated ± 2.4 pp, low confidence
42Qwen3.8 Max Preview46.3%estimated ± 2.7 pp, low confidence
43Kimi K346.0%measured
44Ripple-65B Preview45.8%estimated ± 3.3 pp, medium confidence
45GPT-6.1 Sol45.7%estimated ± 3.3 pp, medium confidence
46Claude Haiku 5.545.5%estimated ± 4.1 pp, high confidence
47Qwen3.8-Flash-Next45.4%measured
48GPT-5.6 Sol44.3%measured
49GPT-5.5 Pro43.9%estimated ± 3.6 pp, medium confidence
50Gemini 3.7 Flash43.4%estimated ± 3.3 pp, medium confidence
51GPT-5.4 Pro43.2%estimated ± 3.6 pp, medium confidence
52Grok 4.542.1%measured
53LFM2.5-24B-A2B42.0%estimated ± 3.3 pp, medium confidence
54Qwen3.8 Max41.8%estimated ± 3.3 pp, medium confidence
55Claude Mythos 541.5%estimated ± 3.3 pp, medium confidence
56Muse Spark Open41.3%estimated ± 4.1 pp, high confidence
57Muse Spark 1.241.2%estimated ± 3.7 pp, high confidence
58Claude Opus 541.1%estimated ± 3.3 pp, medium confidence
59Ornith-1.5-397B40.9%estimated ± 3.6 pp, medium confidence
60Inkling 2 Small40.4%measured
61GPT-5.6 Terra40.2%measured
62Gemini 3.5 Flash Cyber40.0%estimated ± 3.3 pp, medium confidence
63Claude Mythos Preview39.9%estimated ± 3.3 pp, medium confidence
64Gemini 3.5 Flash39.8%estimated ± 2.0 pp, medium confidence
65DeepSeek V4 Pro39.6%measured
66GPT-5.439.6%measured
67GPT-5.6 Luna39.6%estimated ± 2.0 pp, medium confidence
68Kimi K2.639.6%estimated ± 2.0 pp, medium confidence
69Step 3.7 Flash39.6%estimated ± 2.0 pp, medium confidence
70GLM-5.239.6%estimated ± 2.0 pp, medium confidence
71MiniMax M2.739.6%estimated ± 2.0 pp, low confidence
72Claude Opus 4.539.6%estimated ± 2.0 pp, low confidence
73GPT-5.4 mini39.6%estimated ± 2.0 pp, low confidence
74GLM-539.6%estimated ± 2.0 pp, low confidence
75Qwen3.6 Plus39.6%estimated ± 2.0 pp, low confidence
76MiniMax M3.1 Flash39.5%estimated ± 3.3 pp, medium confidence
77DeepSeek V4 Flash 073139.4%measured
78Hy4 preview39.4%estimated ± 3.3 pp, medium confidence
79Claude Opus 4.739.4%estimated ± 3.3 pp, low confidence
80Claude Opus 4.6 (Adaptive)39.2%estimated ± 2.7 pp, medium confidence
81Gemini 3.1 Pro39.2%estimated ± 2.7 pp, medium confidence
82Apodex 1.139.2%estimated ± 2.7 pp, medium confidence
83Apodex 1.1 Mini39.2%estimated ± 2.7 pp, medium confidence
84GPT-5.4 nano39.2%estimated ± 2.7 pp, medium confidence
85Qwen3.7 Plus39.2%estimated ± 2.7 pp, low confidence
86Grok 4.339.2%estimated ± 2.7 pp, low confidence
87Kimi K2.539.2%estimated ± 2.7 pp, low confidence
88Qwen3.5 397B39.2%estimated ± 2.7 pp, low confidence
89GPT-5.539.0%measured
90Ling 3.0 Flash Fin38.6%measured
91Claude Opus 4.638.5%estimated ± 3.6 pp, medium confidence
92dots3-note Preview38.2%estimated ± 3.6 pp, medium confidence
93Claude Fable 538.1%measured
94Claude Sonnet 537.3%measured
95Sage-A1.537.0%estimated ± 3.6 pp, medium confidence
96Ling 3.0 Flash Sante35.8%estimated ± 3.6 pp, medium confidence
97Ling 3.0 Flash VL35.2%estimated ± 4.1 pp, high confidence
98Inkling 2 Mini34.9%estimated ± 3.6 pp, medium confidence
99Quasar 438B34.7%estimated ± 3.7 pp, high confidence
100Nemotron 3.5 Super VL 120B A12B34.2%estimated ± 3.6 pp, medium confidence
101Nemotron 3.5 Super VL 120B A12B NVFP433.9%estimated ± 3.6 pp, medium confidence
102Inkling-Small33.8%estimated ± 3.6 pp, medium confidence
103Solar Pro 433.6%estimated ± 4.1 pp, high confidence
104Inkling33.6%estimated ± 3.6 pp, medium confidence
105Gemini 3.6 Flash33.3%estimated ± 3.7 pp, high confidence
106Beam 1.0 501B-A23B FP833.3%estimated ± 3.6 pp, medium confidence
107Beam 1.0 501B-A23B NVFP432.5%estimated ± 3.6 pp, medium confidence
108Agents-A132.5%estimated ± 3.6 pp, medium confidence
109Muse Spark 1.131.8%estimated ± 3.7 pp, high confidence
110Solar Open 231.7%estimated ± 4.5 pp, low confidence
111Hy330.7%estimated ± 3.7 pp, high confidence
112Hy3 Preview30.7%estimated ± 3.7 pp, high confidence
113GLM-4.730.5%estimated ± 4.1 pp, high confidence
114Muse Spark29.9%estimated ± 4.1 pp, high confidence
115Qwen3.7 Max29.7%estimated ± 3.7 pp, high confidence
116MiMo-V2.5-Pro29.0%estimated ± 3.7 pp, high confidence
117Kimi K2.7 Code28.9%estimated ± 3.7 pp, high confidence
118Nemotron 3.5 Bolt 74B-A9B BF1628.8%estimated ± 3.6 pp, low confidence
119A.X K228.4%estimated ± 4.1 pp, high confidence
120Ling 3.0 Flash28.0%measured
121GLM-5.127.6%estimated ± 3.6 pp, low confidence
122Qwen3.6-27B27.5%estimated ± 3.7 pp, medium confidence
123Ornith-1.5-35B-A3B27.3%estimated ± 3.6 pp, low confidence
124Agents-A1-4B26.8%estimated ± 3.6 pp, low confidence
125Nemotron 3.5 Bolt 74B-A9B26.8%estimated ± 3.6 pp, low confidence
126GPT-526.7%estimated ± 4.1 pp, medium confidence
127GPT-5.226.2%estimated ± 3.6 pp, low confidence
128LLaDA2.2-flash26.0%estimated ± 4.5 pp, low confidence
129Gemini 3.5 Flash-Lite25.2%estimated ± 3.7 pp, medium confidence
130Qwen3.5-122B-A10B25.1%estimated ± 3.6 pp, low confidence
131Qwen3.6-35B-A3B24.6%estimated ± 3.7 pp, medium confidence
132Qwen3.5-27B23.5%estimated ± 3.6 pp, low confidence
133Qwen3.5-35B-A3B23.5%estimated ± 3.6 pp, low confidence
134GPT-5.122.6%estimated ± 4.1 pp, medium confidence
135Muse Glimmer 30B22.0%estimated ± 3.7 pp, medium confidence
136K-EXAONE 2.021.7%estimated ± 4.1 pp, medium confidence
137Mistral Medium 3.5 128B21.4%estimated ± 3.7 pp, medium confidence
138Ornith-1.5-9B21.0%estimated ± 3.6 pp, low confidence
139Gemma 4 31B19.9%estimated ± 3.7 pp, medium confidence
140GPT-OSS 120B19.6%estimated ± 3.7 pp, medium confidence
141Nemotron 3.5 Lightning 30B A3B NVFP419.5%estimated ± 3.7 pp, medium confidence
142Spark-X2.6-4B19.3%estimated ± 3.6 pp, low confidence
143Granite 4.2 8B18.1%estimated ± 3.7 pp, medium confidence
144Command A+18.1%estimated ± 3.7 pp, medium confidence
145Gemini 2.5 Pro18.1%estimated ± 3.7 pp, medium confidence
146Mistral Large 317.4%estimated ± 3.7 pp, medium confidence
147Mistral Small 416.8%estimated ± 3.7 pp, medium confidence
148GPT-OSS 20B16.8%estimated ± 3.7 pp, medium confidence
149Trinity-Large-Preview16.7%estimated ± 3.7 pp, medium confidence
150Trinity-Large-Thinking16.7%estimated ± 3.7 pp, medium confidence
151Nemotron 3 Nano 30B16.6%estimated ± 3.7 pp, medium confidence
152DeepSeek V316.5%estimated ± 3.7 pp, medium confidence
153Celeris-116.4%estimated ± 3.7 pp, medium confidence
154Llama 4 Maverick16.4%estimated ± 3.7 pp, medium confidence
155Llama 4 Scout16.3%estimated ± 3.7 pp, medium confidence
156Gemma 3 27B16.1%estimated ± 3.7 pp, medium confidence
157MiniCPM5-2B16.1%estimated ± 4.1 pp, medium confidence
158MiMo-V2-Flash10.7%estimated ± 4.1 pp, medium confidence
159Granite 4.2 30B6.6%estimated ± 4.1 pp, medium confidence
160Gemma 4 26B A4B5.8%estimated ± 4.1 pp, medium confidence
161Ling 3.0 Tiny5.4%estimated ± 4.1 pp, medium confidence
162Claude 3.5 Haiku0.0%estimated ± 4.1 pp, medium confidence
163DeepSeek V3 03240.0%estimated ± 4.1 pp, medium confidence
164Gemma 4 12B0.0%estimated ± 4.1 pp, medium confidence
165Gemma 4 E2B0.0%estimated ± 4.1 pp, medium confidence
166Gemma 4 E4B0.0%estimated ± 4.1 pp, medium confidence
167GPT-4.1 mini0.0%estimated ± 4.1 pp, medium confidence
168GPT-4.1 nano0.0%estimated ± 4.1 pp, medium confidence
169GPT-4o0.0%estimated ± 4.1 pp, medium confidence
170GPT-4o mini0.0%estimated ± 4.1 pp, medium confidence
171Granite 4.2 3B0.0%estimated ± 4.1 pp, medium confidence
172K-Exaone0.0%estimated ± 4.1 pp, medium confidence
173LFM2.5-2.6B0.0%estimated ± 4.1 pp, medium confidence
174Ling 2.6 Flash0.0%estimated ± 4.1 pp, medium confidence
175Llama 3.3 70B0.0%estimated ± 4.1 pp, medium confidence
176Mercury 2.50.0%estimated ± 4.1 pp, medium confidence
177Mistral Small 3.10.0%estimated ± 4.1 pp, medium confidence
178Nemotron 3 Nano Omni 30B A3B0.0%estimated ± 4.1 pp, medium confidence
179North Mini Code0.0%estimated ± 4.1 pp, medium confidence
180Solar Pro 30.0%estimated ± 4.1 pp, medium confidence
181Ultravox v0.6 Llama 3.3 70B0.0%estimated ± 4.1 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General