Agentic · tools

AA ITBench leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on AA ITBench is 56.2% by GPT-5.6 Sol. 159 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Exaone 4.0 32B	100.0%	estimated ± 3.6 pp, low confidence
2	Phi-4	100.0%	estimated ± 3.6 pp, low confidence
3	LFM2.5-VL-1.6B-Extract	100.0%	estimated ± 3.6 pp, low confidence
4	Gemma 3 27B	100.0%	estimated ± 3.6 pp, low confidence
5	Nemotron Ultra 253B	100.0%	estimated ± 3.6 pp, low confidence
6	Granite-4.0-350M	100.0%	estimated ± 3.6 pp, low confidence
7	Nova Pro	100.0%	estimated ± 3.6 pp, low confidence
8	Granite-4.0-H-350M	100.0%	estimated ± 3.6 pp, low confidence
9	Gemini 2.5 Flash	100.0%	estimated ± 3.6 pp, low confidence
10	Llama 4 Scout	100.0%	estimated ± 3.6 pp, low confidence
11	Qwen3-Omni-30B-A3B-Instruct	100.0%	estimated ± 3.6 pp, low confidence
12	GPT-4.1 nano	100.0%	estimated ± 3.6 pp, low confidence
13	Llama 4 Maverick	100.0%	estimated ± 3.6 pp, low confidence
14	Llama 3.1 405B	100.0%	estimated ± 3.6 pp, low confidence
15	Granite-4.0-H-1B	100.0%	estimated ± 3.6 pp, low confidence
16	Exaone 4.0 1.2B	100.0%	estimated ± 3.6 pp, low confidence
17	Gemma 4 E2B	100.0%	estimated ± 3.6 pp, low confidence
18	Gemma 4 E4B	100.0%	estimated ± 3.6 pp, low confidence
19	Claude 3 Haiku	100.0%	estimated ± 3.6 pp, low confidence
20	Qwen3-Omni-30B-A3B-Thinking	100.0%	estimated ± 3.6 pp, low confidence
21	DeepSeek V3	99.9%	estimated ± 3.6 pp, low confidence
22	Mistral Medium 3	99.9%	estimated ± 3.6 pp, low confidence
23	Mistral Large 3	99.9%	estimated ± 3.6 pp, low confidence
24	GPT-4o	99.9%	estimated ± 3.6 pp, low confidence
25	Ultravox v0.6 Llama 3.3 70B	99.8%	estimated ± 3.6 pp, low confidence
26	o3-mini	99.8%	estimated ± 3.6 pp, low confidence
27	Mistral Large 2	99.6%	estimated ± 3.6 pp, low confidence
28	Solar Pro 2	99.5%	estimated ± 3.6 pp, low confidence
29	Sarvam 30B	99.3%	estimated ± 3.6 pp, low confidence
30	DeepSeek V3.1	99.2%	estimated ± 3.6 pp, low confidence
31	Gemma 4 12B	99.0%	estimated ± 3.6 pp, low confidence
32	DeepSeek-R1	99.0%	estimated ± 3.6 pp, low confidence
33	DeepSeek V3.1 (Reasoning)	98.8%	estimated ± 3.6 pp, low confidence
34	North Mini Code	98.8%	estimated ± 3.6 pp, low confidence
35	Nemotron 3 Nano 30B	98.0%	estimated ± 3.6 pp, low confidence
36	Mistral Small 4	97.9%	estimated ± 3.6 pp, low confidence
37	Mistral Small 4 (Reasoning)	97.9%	estimated ± 3.6 pp, low confidence
38	Gemini 3 Flash	97.3%	estimated ± 3.6 pp, low confidence
39	Gemma 4 26B A4B	97.1%	estimated ± 3.6 pp, low confidence
40	Nemotron 3 Nano Omni 30B A3B	96.4%	estimated ± 3.6 pp, low confidence
41	GLM-4.5-Air	95.9%	estimated ± 3.6 pp, low confidence
42	Sarvam 105B	95.7%	estimated ± 3.6 pp, low confidence
43	DeepSeek V3 0324	95.6%	estimated ± 3.6 pp, low confidence
44	GPT-4.1	95.6%	estimated ± 3.6 pp, low confidence
45	Claude 4 Sonnet	92.2%	estimated ± 3.6 pp, low confidence
46	GPT-4.1 mini	91.7%	estimated ± 3.6 pp, low confidence
47	Gemini 2.5 Pro	90.7%	estimated ± 3.6 pp, low confidence
48	LLaDA2.2-mini	87.3%	estimated ± 3.6 pp, low confidence
49	Gemma 4 31B	84.6%	estimated ± 3.6 pp, low confidence
50	GPT-OSS 20B	84.3%	estimated ± 3.6 pp, low confidence
51	Kimi K2	83.2%	estimated ± 3.6 pp, low confidence
52	o1	81.3%	estimated ± 3.6 pp, low confidence
53	Grok 4.1 Fast	79.9%	estimated ± 3.6 pp, low confidence
54	GPT-OSS 120B	77.1%	estimated ± 3.6 pp, low confidence
55	Grok 4 Fast (Reasoning)	77.1%	estimated ± 3.6 pp, low confidence
56	Nemotron 3 Super 100B	74.3%	estimated ± 3.6 pp, low confidence
57	Ornith-1.5-9B	73.2%	estimated ± 3.4 pp, low confidence
58	Ornith-1.5-35B-A3B	70.7%	estimated ± 3.4 pp, low confidence
59	Claude 4.1 Opus Thinking	69.3%	estimated ± 3.6 pp, low confidence
60	Nemotron 3 Ultra	67.2%	estimated ± 3.4 pp, low confidence
61	Claude Opus 4.7	65.8%	estimated ± 3.6 pp, low confidence
62	K-Exaone	65.4%	estimated ± 3.6 pp, low confidence
63	Qwen3 Max	65.4%	estimated ± 3.6 pp, low confidence
64	Grok 4	64.6%	estimated ± 3.6 pp, low confidence
65	Grok Code Fast 1	63.6%	estimated ± 3.6 pp, low confidence
66	GLM-4.6	62.1%	estimated ± 3.6 pp, low confidence
67	Agents-A1-4B	60.5%	estimated ± 3.6 pp, low confidence
68	DeepSeek V4 Flash 0731	60.4%	estimated ± 3.4 pp, low confidence
69	DeepSeek V3.2	59.6%	estimated ± 3.6 pp, low confidence
70	Claude Sonnet 4.6	58.9%	estimated ± 3.6 pp, low confidence
71	Step 3.7 Flash	58.6%	estimated ± 3.4 pp, low confidence
72	Agents-A1	58.2%	estimated ± 3.4 pp, low confidence
73	Ternary Bonsai 2 27B	58.1%	estimated ± 3.6 pp, low confidence
74	o3	57.6%	estimated ± 3.6 pp, low confidence
75	GPT-5.1	56.3%	estimated ± 3.6 pp, low confidence
76	Muse Spark 1.1	56.3%	estimated ± 3.6 pp, low confidence
77	GPT-5.6 Sol	56.2%	measured
78	Step 5 Preview	55.6%	measured
79	GPT-5.1-Codex	55.1%	estimated ± 3.6 pp, low confidence
80	GPT-5.1-Codex-Max	55.1%	estimated ± 3.6 pp, low confidence
81	MiMo-V2-Flash	54.2%	estimated ± 3.6 pp, low confidence
82	Qwen3.5 397B (Reasoning)	54.2%	estimated ± 3.6 pp, low confidence
83	dots3-note Preview	53.8%	estimated ± 3.4 pp, low confidence
84	Claude Opus 4.6	53.3%	estimated ± 3.6 pp, low confidence
85	GPT-5.2	53.3%	estimated ± 3.6 pp, low confidence
86	GPT-5 (high)	53.3%	estimated ± 3.6 pp, low confidence
87	MiniMax M2.7	53.3%	estimated ± 3.6 pp, low confidence
88	Command A+	53.1%	estimated ± 3.6 pp, low confidence
89	Qwen3.7 Max	53.1%	estimated ± 3.4 pp, low confidence
90	Gemini 3.8 Flash	52.5%	measured
91	GPT-6.1 Sol	52.4%	estimated ± 3.9 pp, medium confidence
92	GPT-5.3 Codex	52.2%	estimated ± 3.6 pp, medium confidence
93	Ling 2.6 Flash	52.2%	estimated ± 3.6 pp, medium confidence
94	Solar Pro 3	52.0%	estimated ± 3.6 pp, medium confidence
95	GPT-5 (medium)	51.8%	estimated ± 3.6 pp, medium confidence
96	Hy4 preview	51.4%	estimated ± 3.4 pp, medium confidence
97	Gemini 3.5 Flash	51.3%	estimated ± 3.6 pp, medium confidence
98	Gemini 3 Pro	51.3%	estimated ± 3.6 pp, medium confidence
99	GLM-5.3-Flash	51.2%	measured
100	GPT-5.6 Terra	51.0%	measured
101	Apodex 1.1	50.8%	estimated ± 3.4 pp, medium confidence
102	Ornith-1.5-397B	50.8%	estimated ± 3.4 pp, medium confidence
103	Qwen3.8 Max	50.7%	estimated ± 3.4 pp, medium confidence
104	LFM2.5-8B-A1B	50.4%	estimated ± 3.6 pp, medium confidence
105	Qwen3.8-27B	50.0%	estimated ± 5.7 pp, low confidence
106	Claude Opus 4.8	49.9%	estimated ± 3.6 pp, medium confidence
107	Claude Haiku 5.5	49.6%	estimated ± 3.4 pp, medium confidence
108	Claude Sonnet 5	49.6%	estimated ± 3.4 pp, medium confidence
109	Qwen3.5-35B-A3B	49.5%	estimated ± 3.6 pp, medium confidence
110	Claude Fable 5.1	49.5%	measured
111	GPT-6 Sol	49.4%	measured
112	Claude Opus 4.5 Thinking	49.3%	estimated ± 3.6 pp, medium confidence
113	Trinity-Large-Preview	48.8%	estimated ± 3.6 pp, medium confidence
114	Trinity-Large-Thinking	48.8%	estimated ± 3.6 pp, medium confidence
115	GPT-6 Astra	48.6%	measured
116	MiMo-V2-Omni	48.0%	estimated ± 3.6 pp, medium confidence
117	Muse Spark	47.8%	estimated ± 3.6 pp, medium confidence
118	Kimi K3	47.7%	measured
119	Claude Opus 4.6 (Adaptive)	47.4%	estimated ± 3.6 pp, medium confidence
120	GPT-5.2-Codex	47.4%	estimated ± 3.6 pp, medium confidence
121	DeepSeek V4 Pro 0813	47.4%	estimated ± 3.4 pp, medium confidence
122	Inkling-Small	47.2%	estimated ± 3.6 pp, medium confidence
123	DeepSeek V4.1 Flash	46.9%	measured
124	Claude Opus 4.7 (Adaptive)	46.7%	measured
125	Grok 4.1 Fast (Reasoning)	46.6%	estimated ± 3.6 pp, medium confidence
126	Qwen3.5-122B-A10B	46.4%	estimated ± 3.6 pp, medium confidence
127	Beam	46.4%	estimated ± 3.6 pp, medium confidence
128	MiMo-V2.6-Pro	46.3%	estimated ± 3.9 pp, medium confidence
129	Claude Mythos Preview	46.3%	estimated ± 3.9 pp, medium confidence
130	GPT-5.6 Luna	46.3%	estimated ± 3.9 pp, low confidence
131	GPT-6 Luna	46.3%	estimated ± 3.9 pp, low confidence
132	MiMo-V2.6-Flash	46.3%	estimated ± 3.9 pp, low confidence
133	Qwen3.5-27B	46.2%	estimated ± 3.6 pp, medium confidence
134	GLM-5.3	46.1%	measured
135	MiMo-V2.5-Pro	46.1%	estimated ± 3.6 pp, medium confidence
136	Mistral Medium 3.5 128B	46.1%	estimated ± 3.6 pp, medium confidence
137	Qwen3.6-27B	46.1%	estimated ± 3.6 pp, medium confidence
138	GPT-5.5	45.8%	measured
139	MiMo-V2-Pro	45.6%	estimated ± 3.6 pp, medium confidence
140	Gemini 3.1 Pro	45.2%	estimated ± 3.6 pp, medium confidence
141	GLM-4.7	45.1%	estimated ± 3.6 pp, medium confidence
142	Kimi K2.5 (Reasoning)	45.1%	estimated ± 3.6 pp, medium confidence
143	Qwen 3.6 Max (preview)	45.1%	estimated ± 3.6 pp, medium confidence
144	Grok 4.3	44.1%	estimated ± 3.6 pp, medium confidence
145	Gemini 3.7 Flash	44.0%	estimated ± 6.6 pp, low confidence
146	Kimi K2.7 Code	43.8%	estimated ± 3.6 pp, medium confidence
147	Claude Fable 5	43.7%	estimated ± 3.6 pp, medium confidence
148	GLM-5-Turbo	43.7%	estimated ± 3.6 pp, medium confidence
149	GLM-5V-Turbo	43.7%	estimated ± 3.6 pp, medium confidence
150	Claude Sonnet 5.5	43.4%	estimated ± 3.4 pp, medium confidence
151	Muse Glimmer 30B	43.3%	estimated ± 3.6 pp, medium confidence
152	Claude Opus 5	43.2%	estimated ± 3.4 pp, medium confidence
153	GLM-5.2	42.7%	measured
154	MiniMax M3	42.1%	estimated ± 3.6 pp, low confidence
155	Grok 4.7	42.1%	measured
156	Inkling	42.1%	estimated ± 3.6 pp, low confidence
157	Qwen3.7 Plus	41.3%	estimated ± 3.6 pp, low confidence
158	GLM-5.1	40.0%	estimated ± 3.6 pp, low confidence
159	GPT-5.4	39.0%	estimated ± 3.6 pp, low confidence
160	Claude Opus 5.5	38.2%	measured
161	Ling 3.0 Flash	34.8%	estimated ± 3.6 pp, low confidence
162	Grok 4.6	33.3%	estimated ± 5.7 pp, low confidence
163	Muse Spark 1.3	33.2%	measured
164	Qwen3.6-35B-A3B	32.7%	estimated ± 3.6 pp, low confidence
165	Solar Pro 4	31.7%	estimated ± 3.6 pp, low confidence
166	Solar Open 2	29.3%	estimated ± 3.6 pp, low confidence
167	GPT-5.4 mini	29.0%	estimated ± 3.6 pp, low confidence
168	GPT-5.4 nano	27.9%	estimated ± 3.6 pp, low confidence
169	Kimi K2.6	27.7%	estimated ± 3.6 pp, low confidence
170	Qwen3.6 Plus	22.7%	estimated ± 3.6 pp, low confidence
171	LLaDA2.2-flash	21.5%	estimated ± 3.6 pp, low confidence
172	Qwen3.5 397B	21.4%	estimated ± 3.6 pp, low confidence
173	LongCat-Flash-Lite-Sparse	21.1%	estimated ± 3.6 pp, low confidence
174	Claude Opus 4.5	19.2%	estimated ± 3.6 pp, low confidence
175	GLM-5	13.2%	estimated ± 3.6 pp, low confidence
176	Kimi K2.5	12.4%	estimated ± 3.6 pp, low confidence

AA ITBench leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General