Knowledge & reasoning

Artificial Analysis Intelligence Index leaderboard

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

As of 2026-10-07, the highest measured score on Artificial Analysis Intelligence Index is 58.9% by GPT-5.6 Sol. 45 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	GPT-5.6 Sol	58.9%	measured
2	Claude Opus 5.5	57.6%	measured
3	Claude Sonnet 5.5	56.0%	measured
4	GPT-5.6 Terra	54.9%	measured
5	Claude Fable 5.1	53.3%	measured
6	DeepSeek V4 Pro 0813	53.2%	measured
7	GPT-6 Astra	52.7%	measured
8	Gemini 4 Argon	52.6%	measured
9	GPT-6.1 Sol	51.8%	measured
10	Claude Mythos 5	51.5%	estimated ± 5.1 pp, medium confidence
11	GPT-5.6 Luna	51.2%	measured
12	Claude Opus 5	50.8%	measured
13	Gemini 3.5 Flash	50.2%	measured
14	GPT-5.5 Pro	49.7%	estimated ± 5.1 pp, medium confidence
15	Claude Fable 5	49.6%	measured
16	GPT-5.4 Pro	49.4%	estimated ± 5.1 pp, medium confidence
17	Muse Spark 1.3	48.1%	measured
18	GPT-6 Sol	47.6%	measured
19	Gemini 3 Pro Deep Think	46.8%	estimated ± 5.1 pp, medium confidence
20	Grok 4.7	46.5%	measured
21	MiMo-V2.6-Pro	46.3%	measured
22	dots3-note Preview	45.8%	estimated ± 7.0 pp, medium confidence
23	Qwen3.8 Max Preview	45.4%	measured
24	GLM-5.3	44.8%	measured
25	Grok 4.6	44.3%	measured
26	Step 5 Preview	43.7%	measured
27	Kimi K3	43.6%	measured
28	Claude Haiku 5.5	43.4%	measured
29	Sakana Fugu-Ultra	43.3%	estimated ± 5.1 pp, medium confidence
30	Pareto 26.9	42.4%	estimated ± 5.1 pp, medium confidence
31	GLM-5.3-Flash	41.8%	measured
32	Claude Opus 4.8	41.8%	measured
33	Hy3 Preview	41.2%	measured
34	Ling 3.1 Flash	41.1%	measured
35	Gemini 3.8 Flash	40.9%	measured
36	Sakana Fugu	40.7%	estimated ± 5.1 pp, medium confidence
37	Claude Opus 4.7 (Adaptive)	40.7%	measured
38	Hy4 preview	40.4%	estimated ± 5.1 pp, medium confidence
39	Qwen3.8-Flash-Next	39.8%	measured
40	Muse Spark 1.2	39.6%	measured
41	Pareto 26.10 Preview	39.6%	estimated ± 7.4 pp, medium confidence
42	DeepSeek V4.1 Flash	39.5%	measured
43	Gemini 3.7 Flash	39.1%	measured
44	GPT-5.4	39.0%	measured
45	Grok 4.5	38.8%	measured
46	Mistral Large 4	38.4%	measured
47	GPT-5.5	38.4%	measured
48	Ornith-1.5-397B	38.3%	estimated ± 5.1 pp, medium confidence
49	Claude Sonnet 5	38.2%	measured
50	GPT-6 Luna	38.1%	measured
51	MiMo-V2.6-Flash	37.9%	measured
52	Grok 4.3	37.6%	measured
53	Qwen3.8 Max	37.4%	estimated ± 5.1 pp, medium confidence
54	Qwen3.8-Omni-Flash	34.9%	estimated ± 6.2 pp, medium confidence
55	DeepSeek V4 Flash 0731	34.3%	measured
56	Agents-A1	34.1%	estimated ± 8.7 pp, low confidence
57	Gemini 3.6 Flash	34.0%	measured
58	Muse Spark 1.1	33.7%	measured
59	GLM-5.2	33.7%	measured
60	Qwen3.8-27B	33.7%	measured
61	GPT-5.3 Codex	32.5%	measured
62	Claude Opus 4.6 (Adaptive)	32.0%	measured
63	Interfaze Beta	31.5%	estimated ± 6.2 pp, medium confidence
64	Muse Spark	31.3%	measured
65	Claude Opus 4.7	30.9%	measured
66	Beam	30.8%	estimated ± 5.1 pp, medium confidence
67	GPT-5.2	30.5%	measured
68	Gemini 3.1 Pro	29.7%	measured
69	Qwen3.7 Max	29.5%	measured
70	MiniMax M3	29.2%	measured
71	Claude Opus 4.5 Thinking	29.1%	measured
72	MiMo-V2-Pro	28.6%	measured
73	GPT-5.2-Codex	28.5%	measured
74	Qwen 3.6 Max (preview)	28.4%	measured
75	Solar Pro 4	28.2%	measured
76	Gemini 3 Pro	28.0%	measured
77	GLM-5	27.9%	measured
78	Qwen3.6 Plus	27.0%	measured
79	Kimi K2.6	27.0%	measured
80	Grok 4.20	26.9%	estimated ± 5.1 pp, medium confidence
81	Quasar 438B	26.7%	measured
82	GLM-5-Turbo	26.6%	measured
83	Apodex 1.1	26.4%	measured
84	Apodex 1.1 Mini	26.4%	measured
85	Claude Opus 4.6	26.4%	measured
86	GLM-5.1	26.1%	measured
87	MiMo-V2.5-Pro	26.0%	measured
88	Kimi K2.7 Code	25.8%	measured
89	Inkling-Small	25.7%	measured
90	Hy3	25.3%	measured
91	Qwen3.7 Plus	25.2%	measured
92	Inkling	25.0%	measured
93	Solar Open 2	24.8%	estimated ± 5.1 pp, medium confidence
94	GPT-5.1	24.7%	measured
95	Claude Sonnet 4.6	24.7%	measured
96	Ling 3.0 Flash VL	24.6%	measured
97	GPT-5.4 mini	24.1%	measured
98	MiMo-V2-Omni	23.9%	measured
99	GPT-5.1-Codex	23.7%	measured
100	GPT-5.1-Codex-Max	23.7%	measured
101	Claude Opus 4.5	23.7%	measured
102	GLM-5V-Turbo	23.5%	measured
103	Kimi K2.5	23.5%	measured
104	Kimi K2.5 (Reasoning)	23.5%	measured
105	Seed 2.1 Pro	23.3%	estimated ± 3.9 pp, medium confidence
106	GPT-5 (high)	23.0%	measured
107	Nemotron 3 Ultra	22.9%	measured
108	Qwen3.5-27B	22.9%	measured
109	GPT-5 (medium)	22.9%	measured
110	Claude 4.1 Opus Thinking	22.9%	measured
111	MiniMax M2.7	22.8%	measured
112	Command A+	22.5%	measured
113	Grok 4	22.5%	measured
114	Seed 2.1 Turbo	22.5%	estimated ± 3.9 pp, medium confidence
115	Ornith-1.5-35B-A3B	22.4%	estimated ± 5.1 pp, medium confidence
116	Qwen3.5 Flash	22.3%	estimated ± 7.9 pp, medium confidence
117	GLM-4.7	22.2%	measured
118	Gemini 3.5 Flash-Lite	22.2%	measured
119	o3-pro	21.9%	measured
120	Claude Sonnet 4.5 Thinking	21.7%	estimated ± 7.0 pp, medium confidence
121	A.X K2	21.5%	measured
122	Qwen3.5 397B	21.4%	measured
123	Qwen3.5 397B (Reasoning)	21.4%	measured
124	Qwen3.6-27B	21.4%	measured
125	Qwen3 235B 2507	21.3%	estimated ± 3.9 pp, medium confidence
126	MiMo-V2.5	20.9%	estimated ± 7.9 pp, medium confidence
127	GPT-5.4 nano	20.7%	measured
128	Gemini 3.1 Flash-Lite	20.5%	estimated ± 7.9 pp, medium confidence
129	Grok 4.1 Fast (Reasoning)	20.4%	measured
130	o3	20.2%	measured
131	Ling 3.0 Flash	20.1%	measured
132	Ling 3.0 Flash FP8	20.1%	measured
133	K-EXAONE 2.0	19.7%	measured
134	Step 3.7 Flash	19.5%	measured
135	Qwen3.5-35B-A3B	19.3%	measured
136	MAI-Thinking-1	19.2%	estimated ± 6.2 pp, medium confidence
137	Ornith-1.5-9B	19.0%	estimated ± 5.1 pp, medium confidence
138	Claude 4.1 Opus	18.6%	measured
139	Qwen3.6-35B-A3B	18.2%	measured
140	Claude Sonnet 4.5	18.1%	estimated ± 6.2 pp, medium confidence
141	Ternary Bonsai 2 27B	18.0%	estimated ± 4.2 pp, medium confidence
142	Grok 4 Fast (Reasoning)	17.9%	measured
143	Gemini 3 Flash	17.9%	measured
144	Muse Glimmer 30B	17.5%	measured
145	Claude Haiku 4.5	17.2%	estimated ± 7.9 pp, medium confidence
146	GLM-4.5	17.2%	estimated ± 7.9 pp, medium confidence
147	Laguna XS.2	16.8%	estimated ± 7.9 pp, medium confidence
148	Laguna M.1	16.8%	estimated ± 7.9 pp, low confidence
149	Gemma 4 26B A4B	16.7%	measured
150	Claude 4 Sonnet	16.6%	measured
151	Gemini 2.5 Pro	16.1%	measured
152	MiMo-V2-Flash	16.1%	measured
153	DeepSeek V3.2	16.0%	measured
154	Qwen3 Max	15.6%	measured
155	Qwen3.5-122B-A10B	15.6%	measured
156	o1	15.2%	measured
157	Mellum2-12B-A2.5B-Thinking	15.2%	estimated ± 4.2 pp, medium confidence
158	GLM-4.6	14.9%	measured
159	Granite 4.2 30B	14.8%	measured
160	Gemma 4 31B	14.7%	measured
161	K-Exaone	14.4%	measured
162	Mistral Medium 3.5 128B	14.2%	measured
163	Gemma 4 12B	14.2%	measured
164	Grok Code Fast 1	14.1%	measured
165	Ling 2.6 Flash	14.1%	measured
166	DeepSeek V3.1	13.7%	measured
167	DeepSeek V3.1 (Reasoning)	13.5%	measured
168	DeepSeek-R1	13.1%	measured
169	Nemotron 3.5 Lightning 30B A3B NVFP4	12.9%	measured
170	Nemotron 3 Super 100B	12.8%	measured
171	Kimi K2	12.7%	measured
172	GPT-4.1	12.7%	measured
173	o3-mini	12.5%	measured
174	MiniCPM5-2B	12.5%	measured
175	o1-pro	12.4%	measured
176	Mercury 2.5	12.3%	measured
177	GPT-OSS 120B	11.6%	measured
178	o1-preview	11.4%	measured
179	Grok 4.1 Fast	11.3%	measured
180	Mistral Small 4	11.3%	measured
181	Mistral Small 4 (Reasoning)	11.3%	measured
182	ZAYA1-8B	11.2%	estimated ± 6.2 pp, medium confidence
183	GLM-4.5-Air	11.1%	measured
184	Granite 4.2 8B	11.1%	measured
185	Ling 3.0 Tiny	11.1%	measured
186	Trinity-Large-Preview	10.8%	measured
187	Trinity-Large-Thinking	10.8%	measured
188	Claude 3.5 Sonnet	10.4%	estimated ± 6.2 pp, medium confidence
189	ZAYA1-74B-Preview	10.3%	estimated ± 6.2 pp, medium confidence
190	Soofi S 30B-A3B	10.3%	estimated ± 6.2 pp, medium confidence
191	LFM2.5-230M	10.3%	estimated ± 6.2 pp, low confidence
192	LFM2.5-VL-450M	10.3%	estimated ± 6.2 pp, low confidence
193	Nemotron 3 Nano Omni 30B A3B	10.3%	measured
194	GPT-4.1 mini	10.2%	measured
195	LongCat-Flash-Lite-Sparse	10.0%	estimated ± 1.3 pp, medium confidence
196	Llama 4 Maverick	10.0%	measured
197	North Mini Code	9.9%	measured
198	Gemini 2.5 Flash	9.9%	measured
199	DeepSeek V3 0324	9.7%	measured
200	MiniCPM5-1B	9.6%	estimated ± 3.9 pp, low confidence
201	Mistral Large 3	9.3%	measured
202	Granite 4.2 3B	9.1%	measured
203	Mistral Medium 3	9.0%	measured
204	GPT-OSS 20B	9.0%	measured
205	Gemma 4 E4B	8.9%	measured
206	Nemotron 3 Nano 30B	8.9%	measured
207	Mellum2-12B-A2.5B-Instruct	8.9%	estimated ± 4.2 pp, low confidence
208	Sarvam 105B	8.8%	measured
209	Claude 3 Opus	8.7%	measured
210	DeepSeek V3	8.5%	measured
211	GPT-4o	8.4%	measured
212	LFM2.5-2.6B	8.4%	measured
213	DeepSeek R1 Distill Qwen 32B	8.4%	measured
214	Llama 4 Scout	8.1%	measured
215	Gemini 1.5 Pro	7.9%	measured
216	GPT-4.1 nano	7.8%	measured
217	Solar Pro 3	7.8%	measured
218	Gemma 4 E2B	7.8%	measured
219	Qwen3-Omni-30B-A3B-Thinking	7.8%	measured
220	Ultravox v0.6 Llama 3.3 70B	7.7%	measured
221	Mistral Large 2	7.6%	measured
222	Nemotron Ultra 253B	7.5%	measured
223	Llama 3.1 405B	7.3%	measured
224	LFM2.5-8B-A1B	7.2%	measured
225	GPT-4 Turbo	7.0%	measured
226	Solar Pro 2	7.0%	measured
227	Nova Pro	7.0%	measured
228	Qwen2.5 Coder 32B Instruct	6.7%	measured
229	GPT-4o mini	6.7%	measured
230	Sarvam 30B	6.6%	measured
231	Celeris-1	6.4%	measured
232	Exaone 4.0 32B	6.3%	measured
233	Qwen3-Omni-30B-A3B-Instruct	6.0%	measured
234	Phi-4	5.9%	measured
235	Phi-4 Multimodal Instruct	5.8%	measured
236	Claude 3 Haiku	5.6%	measured
237	Gemini 1.0 Pro	5.3%	measured
238	Exaone 4.0 1.2B	5.2%	measured
239	Granite-4.0-H-1B	5.2%	measured
240	Gemma 3 27B	4.9%	measured
241	Granite-4.0-350M	4.8%	measured
242	Granite-4.0-H-350M	4.8%	measured
243	LFM2.5-VL-1.6B-Extract	4.8%	measured
244	LLaDA2.2-mini	0.6%	estimated ± 7.4 pp, low confidence

Artificial Analysis Intelligence Index leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General