benchgap
Agentic · tools

AA ITBench leaderboard

As of 2026-11-21, the highest measured score on AA ITBench is 56.2% by GPT-5.6 Sol. 175 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-5.6 Sol56.2%measured
2Step 5 Preview55.6%measured
3Gemini 3.8 Flash52.5%measured
4GLM-5.3-Flash51.2%measured
5Atria Dawn Preview51.1%estimated ± 4.1 pp, medium confidence
6GPT-5.6 Terra51.0%measured
7Kimi K3.1 Pro50.9%estimated ± 4.1 pp, high confidence
8Claude Opus 550.8%estimated ± 4.0 pp, high confidence
9LongCat-2.550.8%estimated ± 4.1 pp, high confidence
10Ling 3.1 Flash50.7%estimated ± 4.1 pp, high confidence
11Muse Spark 1.350.7%estimated ± 4.0 pp, high confidence
12Fugu Cyber50.5%estimated ± 4.9 pp, medium confidence
13Gemini 3.8 Flash Cyber50.2%estimated ± 4.9 pp, medium confidence
14Inkling 2 Small50.1%estimated ± 4.1 pp, high confidence
15GPT-5.5 Pro50.1%estimated ± 4.1 pp, high confidence
16Grok 4.649.9%estimated ± 4.0 pp, high confidence
17GPT-5.4 Pro49.8%estimated ± 4.1 pp, high confidence
18Muse Spark Open49.7%estimated ± 4.4 pp, medium confidence
19Claude Fable 5.149.5%measured
20GPT-6 Sol49.4%measured
21Claude Mythos 549.2%estimated ± 4.1 pp, high confidence
22Claude Fable 549.2%estimated ± 4.0 pp, high confidence
23ERNIE 5.5 Flash Preview49.0%estimated ± 4.1 pp, high confidence
24Gemini 3.5 Flash Cyber48.9%estimated ± 4.9 pp, medium confidence
25Claude Mythos Preview48.9%estimated ± 4.9 pp, medium confidence
26Qwen3.8 Max Preview48.7%estimated ± 4.0 pp, high confidence
27DeepSeek V4 Pro48.7%estimated ± 4.0 pp, high confidence
28Ornith-1.5-397B48.6%estimated ± 4.1 pp, high confidence
29GPT-6 Astra48.6%measured
30MiniMax M3.1 Flash48.5%estimated ± 4.9 pp, medium confidence
31Muse Wonder Full48.2%estimated ± 5.1 pp, low confidence
32Claude Sonnet 5.548.2%estimated ± 5.1 pp, low confidence
33Muse Wonder48.1%estimated ± 5.1 pp, low confidence
34Qwen4 Max47.8%estimated ± 6.3 pp, low confidence
35Qwen3.8-Flash-Next47.8%estimated ± 5.1 pp, low confidence
36Gemini 4 Argon47.7%estimated ± 5.1 pp, low confidence
37Claude Haiku 5.547.7%estimated ± 5.1 pp, low confidence
38Qwen3.8-27B47.7%estimated ± 4.0 pp, high confidence
39Mistral Large 447.7%estimated ± 4.1 pp, high confidence
40Kimi K347.7%measured
41GPT-6.1 Sol47.6%estimated ± 5.1 pp, low confidence
42Hy4 preview47.4%estimated ± 4.4 pp, high confidence
43dots3-note Preview47.2%estimated ± 4.1 pp, high confidence
44Claude Sonnet 547.0%estimated ± 4.0 pp, high confidence
45Muse Spark 1.246.9%estimated ± 4.0 pp, high confidence
46DeepSeek V4.1 Flash 101746.9%measured
47Claude Opus 4.746.7%measured
48GPT-6 Luna46.6%estimated ± 5.1 pp, low confidence
49Claude Opus 4.846.5%estimated ± 2.8 pp, medium confidence
50Sage-A1.546.5%estimated ± 4.1 pp, high confidence
51Grok 4.546.3%estimated ± 4.0 pp, high confidence
52DeepSeek V4 Flash 073146.2%estimated ± 4.0 pp, high confidence
53GLM-5.346.1%measured
54Ling 3.0 Flash Sante45.8%estimated ± 4.1 pp, high confidence
55Qwen3.8 Max45.8%estimated ± 4.7 pp, low confidence
56GPT-5.545.8%measured
57Inkling 2 Mini45.3%estimated ± 4.1 pp, high confidence
58Nemotron 3.5 Super VL 120B A12B44.9%estimated ± 4.1 pp, high confidence
59Nemotron 3.5 Super VL 120B A12B NVFP444.6%estimated ± 4.1 pp, high confidence
60Holo3-35B-A3B44.5%estimated ± 4.7 pp, low confidence
61Gemini 3.7 Flash44.5%estimated ± 4.0 pp, high confidence
62Beam 1.0 501B-A23B44.3%estimated ± 4.1 pp, high confidence
63MiMo-V2.6-Pro44.3%estimated ± 4.7 pp, low confidence
64Beam 1.0 501B-A23B FP844.3%estimated ± 4.1 pp, high confidence
65MiMo-V2.6-Flash43.9%estimated ± 4.7 pp, low confidence
66Beam 1.0 501B-A23B NVFP443.8%estimated ± 4.1 pp, high confidence
67Agents-A143.8%estimated ± 4.1 pp, high confidence
68Quasar 438B43.3%estimated ± 4.0 pp, high confidence
69Holo3-122B-A10B43.1%estimated ± 4.7 pp, low confidence
70MiniMax M342.7%estimated ± 4.0 pp, high confidence
71GLM-5.242.7%measured
72Qwen3.7 Max42.5%measured
73Gemini 3.6 Flash42.5%estimated ± 4.0 pp, high confidence
74UI-Mate-27B42.4%estimated ± 4.7 pp, low confidence
75GPT-5.442.4%estimated ± 2.8 pp, high confidence
76Grok 4.742.1%measured
77Claude Opus 4.542.1%estimated ± 2.8 pp, high confidence
78Ripple-65B Preview42.0%estimated ± 4.4 pp, high confidence
79Apodex 1.141.9%estimated ± 5.1 pp, low confidence
80Apodex 1.1 Mini41.9%estimated ± 5.1 pp, low confidence
81Muse Spark 1.141.7%estimated ± 4.0 pp, high confidence
82Gemini 3 Pro41.6%estimated ± 2.8 pp, high confidence
83Claude Sonnet 4.641.5%estimated ± 2.8 pp, high confidence
84Nemotron 3.5 Bolt 74B-A9B BF1641.2%estimated ± 4.1 pp, high confidence
85Ling 3.0 Flash VL41.1%estimated ± 5.1 pp, low confidence
86Claude Opus 4.641.1%estimated ± 2.8 pp, high confidence
87Hy341.0%estimated ± 4.0 pp, high confidence
88Inkling-Small40.8%estimated ± 4.0 pp, high confidence
89Inkling40.6%estimated ± 4.0 pp, high confidence
90Gemini 3.5 Flash40.3%measured
91GLM-5.140.3%measured
92GPT-5.6 Luna40.3%measured
93Ornith-1.5-35B-A3B40.1%estimated ± 4.1 pp, high confidence
94Kimi K2.7 Code40.0%estimated ± 4.0 pp, high confidence
95Nemotron 3 Ultra39.8%estimated ± 4.0 pp, high confidence
96Agents-A1-4B39.7%estimated ± 4.1 pp, high confidence
97Nemotron 3.5 Bolt 74B-A9B39.7%estimated ± 4.1 pp, high confidence
98Ling 3.0 Flash39.5%estimated ± 4.0 pp, high confidence
99GPT-5.3 Codex39.4%estimated ± 2.8 pp, high confidence
100Gemini 3.1 Pro39.3%estimated ± 2.8 pp, high confidence
101Kimi K2.639.2%estimated ± 2.8 pp, high confidence
102Gemini 3 Flash39.2%estimated ± 2.8 pp, high confidence
103Qwen3.7 Plus39.1%estimated ± 4.0 pp, high confidence
104Claude Opus 4.6 (Adaptive)39.1%estimated ± 6.3 pp, low confidence
105Qwen3.6-27B38.7%estimated ± 2.8 pp, high confidence
106GPT-5.4 nano38.5%estimated ± 4.0 pp, high confidence
107UI-Mate-9B38.2%estimated ± 4.7 pp, low confidence
108Claude Opus 5.538.2%measured
109MiMo-V2.5-Pro38.2%measured
110Gemini 3.5 Flash-Lite37.9%estimated ± 4.0 pp, high confidence
111GPT-5.2-Codex37.9%estimated ± 2.8 pp, high confidence
112Step 3.7 Flash37.9%estimated ± 2.8 pp, high confidence
113GLM-537.7%estimated ± 2.8 pp, high confidence
114Qwen3.6 Plus37.7%estimated ± 2.8 pp, high confidence
115GPT-5.1-Codex37.5%estimated ± 2.8 pp, high confidence
116Grok Build 0.137.4%estimated ± 2.8 pp, high confidence
117Gemma 4 31B37.3%measured
118Claude Sonnet 4.537.3%estimated ± 2.8 pp, high confidence
119Muse Spark37.3%estimated ± 5.1 pp, low confidence
120Ling 3.0 Flash Fin37.1%estimated ± 6.3 pp, low confidence
121Grok 4.1 Fast37.1%estimated ± 2.8 pp, high confidence
122MiMo-V2.537.0%estimated ± 2.8 pp, high confidence
123Qwen3.5 397B37.0%estimated ± 2.8 pp, high confidence
124GPT-5.237.0%estimated ± 2.8 pp, high confidence
125Kimi K2.536.9%estimated ± 2.8 pp, high confidence
126A.X K236.7%estimated ± 5.1 pp, low confidence
127Grok 4.336.7%estimated ± 2.8 pp, high confidence
128Qwen3 Max36.7%estimated ± 2.8 pp, high confidence
129Qwen3.6-35B-A3B36.6%estimated ± 2.8 pp, high confidence
130Grok 436.5%estimated ± 2.8 pp, high confidence
131Gemini 2.5 Pro36.5%estimated ± 2.8 pp, high confidence
132GPT-5.136.4%estimated ± 2.8 pp, high confidence
133MiniMax M2.736.4%estimated ± 2.8 pp, high confidence
134GLM-4.736.4%estimated ± 2.8 pp, high confidence
135Claude 4 Sonnet36.3%estimated ± 2.8 pp, high confidence
136GPT-536.3%estimated ± 5.1 pp, low confidence
137Mistral Medium 3.5 128B36.3%estimated ± 2.8 pp, high confidence
138Gemini 3.1 Flash-Lite36.3%estimated ± 2.8 pp, high confidence
139Grok 4.2036.3%estimated ± 2.8 pp, high confidence
140Hy3 Preview36.2%estimated ± 2.8 pp, high confidence
141Muse Glimmer 30B36.2%estimated ± 4.0 pp, high confidence
142MiMo-V2-Pro36.2%estimated ± 2.8 pp, high confidence
143Trinity-Large-Thinking36.0%estimated ± 2.8 pp, medium confidence
144GLM-5V-Turbo36.0%estimated ± 2.8 pp, medium confidence
145GPT-OSS 120B36.0%estimated ± 2.8 pp, medium confidence
146DeepSeek V3.236.0%estimated ± 2.8 pp, medium confidence
147Qwen3.5-35B-A3B36.0%estimated ± 2.8 pp, medium confidence
148GPT-4.135.9%estimated ± 2.8 pp, medium confidence
149Qwen3.5-122B-A10B35.9%estimated ± 4.0 pp, high confidence
150K-EXAONE 2.035.8%estimated ± 5.1 pp, low confidence
151MiniCPM5-2B35.7%estimated ± 5.1 pp, low confidence
152MiMo-V2-Flash35.7%estimated ± 5.1 pp, low confidence
153Granite 4.2 30B35.7%estimated ± 5.1 pp, low confidence
154Gemma 4 26B A4B35.7%estimated ± 5.1 pp, low confidence
155Ling 3.0 Tiny35.7%estimated ± 5.1 pp, low confidence
156Claude 3.5 Haiku35.7%estimated ± 5.1 pp, low confidence
157DeepSeek V3 032435.7%estimated ± 5.1 pp, low confidence
158Gemma 4 12B35.7%estimated ± 5.1 pp, low confidence
159Gemma 4 E2B35.7%estimated ± 5.1 pp, low confidence
160Gemma 4 E4B35.7%estimated ± 5.1 pp, low confidence
161GPT-4.1 mini35.7%estimated ± 5.1 pp, low confidence
162GPT-4.1 nano35.7%estimated ± 5.1 pp, low confidence
163GPT-4o35.7%estimated ± 5.1 pp, low confidence
164GPT-4o mini35.7%estimated ± 5.1 pp, low confidence
165Granite 4.2 3B35.7%estimated ± 5.1 pp, low confidence
166K-Exaone35.7%estimated ± 5.1 pp, low confidence
167LFM2.5-2.6B35.7%estimated ± 5.1 pp, low confidence
168Ling 2.6 Flash35.7%estimated ± 5.1 pp, low confidence
169Llama 3.3 70B35.7%estimated ± 5.1 pp, low confidence
170Mercury 2.535.7%estimated ± 5.1 pp, low confidence
171Mistral Small 3.135.7%estimated ± 5.1 pp, low confidence
172Nemotron 3 Nano Omni 30B A3B35.7%estimated ± 5.1 pp, low confidence
173North Mini Code35.7%estimated ± 5.1 pp, low confidence
174Solar Pro 335.7%estimated ± 5.1 pp, low confidence
175Ultravox v0.6 Llama 3.3 70B35.7%estimated ± 5.1 pp, low confidence
176Qwen3.5-27B35.5%measured
177Solar Open 235.5%estimated ± 4.4 pp, high confidence
178GPT-5.4 mini35.2%measured
179Nemotron 3.5 Lightning 30B A3B NVFP434.8%estimated ± 4.0 pp, medium confidence
180Ornith-1.5-9B34.6%estimated ± 4.1 pp, medium confidence
181Granite 4.2 8B34.0%estimated ± 4.0 pp, medium confidence
182Command A+34.0%estimated ± 4.0 pp, medium confidence
183Mistral Large 333.6%estimated ± 4.0 pp, medium confidence
184Mistral Small 433.3%estimated ± 4.0 pp, medium confidence
185GPT-OSS 20B33.3%estimated ± 4.0 pp, medium confidence
186Trinity-Large-Preview33.2%estimated ± 4.0 pp, medium confidence
187Nemotron 3 Nano 30B33.1%estimated ± 4.0 pp, medium confidence
188DeepSeek V333.1%estimated ± 4.0 pp, medium confidence
189Celeris-133.0%estimated ± 4.0 pp, medium confidence
190Llama 4 Maverick33.0%estimated ± 4.0 pp, medium confidence
191Llama 4 Scout33.0%estimated ± 4.0 pp, medium confidence
192Spark-X2.6-4B32.9%estimated ± 4.1 pp, medium confidence
193Gemma 3 27B32.9%estimated ± 4.0 pp, medium confidence
194Sarvam 105B31.0%estimated ± 4.1 pp, medium confidence
195Solar Pro 430.9%estimated ± 4.1 pp, medium confidence
196LongCat-Flash-Lite-Sparse30.6%estimated ± 4.1 pp, medium confidence
197LLaDA2.2-flash29.2%estimated ± 4.4 pp, medium confidence
198Sarvam 30B23.3%estimated ± 4.1 pp, medium confidence
199LFM3-1.7B21.5%estimated ± 4.4 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General