Knowledge & reasoning

ARC-AGI-1 leaderboard

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

As of 2026-10-07, the highest measured score on ARC-AGI-1 is 98.5% by Claude Fable 5. 207 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Gemini 4 Argon	99.7%	estimated ± 0.7 pp, low confidence
2	Claude Fable 5	98.5%	measured
3	Gemini 3.8 Flash	98.5%	measured
4	GPT-6 Astra	98.5%	measured
5	Claude Fable 5.1	97.5%	measured
6	Claude Opus 5	97.5%	measured
7	Claude Opus 5.5	97.5%	measured
8	Claude Mythos 5	97.4%	estimated ± 3.7 pp, high confidence
9	Claude Sonnet 5.5	97.2%	estimated ± 3.7 pp, high confidence
10	Muse Spark 1.1	96.8%	estimated ± 3.7 pp, high confidence
11	Sakana Fugu	96.5%	estimated ± 2.7 pp, high confidence
12	Sakana Fugu-Ultra	96.5%	estimated ± 2.7 pp, high confidence
13	GPT-5.6 Sol	96.5%	measured
14	GPT-5.6 Terra	96.5%	measured
15	GPT-6.1 Sol	96.5%	measured
16	Pareto 26.9	96.3%	estimated ± 3.7 pp, high confidence
17	Claude Haiku 5.5	95.5%	estimated ± 3.7 pp, high confidence
18	Gemini 3.7 Flash	95.5%	measured
19	GPT-6 Sol	95.5%	measured
20	Claude Sonnet 5	95.4%	estimated ± 0.7 pp, low confidence
21	dots3-note Preview	95.0%	estimated ± 1.9 pp, high confidence
22	GPT-5.5	95.0%	measured
23	GPT-5.5 Pro	95.0%	measured
24	GPT-5.4 Pro	94.5%	measured
25	Kimi K3	94.5%	measured
26	Muse Spark 1.3	94.4%	estimated ± 5.4 pp, medium confidence
27	Grok 4.7	94.3%	estimated ± 5.4 pp, medium confidence
28	MiMo-V2.6-Pro	94.2%	estimated ± 5.4 pp, medium confidence
29	Qwen3.8 Max Preview	94.1%	estimated ± 5.4 pp, medium confidence
30	GLM-5.3	94.1%	estimated ± 5.4 pp, medium confidence
31	Step 5 Preview	94.1%	estimated ± 2.7 pp, high confidence
32	Gemini 3.1 Pro	94.0%	estimated ± 1.9 pp, high confidence
33	GPT-5.4	93.7%	measured
34	Claude Opus 4.7 (Adaptive)	93.7%	estimated ± 1.9 pp, high confidence
35	GLM-5.3-Flash	93.5%	estimated ± 5.4 pp, medium confidence
36	Ling 3.1 Flash	93.4%	estimated ± 5.4 pp, medium confidence
37	Ornith-1.5-397B	93.1%	estimated ± 2.7 pp, high confidence
38	Claude Opus 4.6 (Adaptive)	93.0%	measured
39	Muse Spark 1.2	93.0%	estimated ± 5.4 pp, medium confidence
40	Qwen3.8 Max	92.9%	estimated ± 2.7 pp, high confidence
41	Mistral Large 4	92.6%	estimated ± 5.4 pp, medium confidence
42	Pareto 26.10 Preview	92.6%	estimated ± 2.7 pp, high confidence
43	Qwen3.7 Max	92.6%	estimated ± 2.7 pp, high confidence
44	Claude Opus 4.8	92.5%	measured
45	Gemini 3.5 Flash	92.5%	measured
46	Hy4 preview	92.4%	estimated ± 2.7 pp, high confidence
47	MiMo-V2.6-Flash	92.4%	estimated ± 5.4 pp, medium confidence
48	Grok 4.3	92.3%	estimated ± 5.4 pp, medium confidence
49	Qwen3.8-Flash-Next	91.6%	estimated ± 2.7 pp, high confidence
50	Gemini 3.6 Flash	91.2%	measured
51	GLM-5.2	90.9%	estimated ± 2.7 pp, high confidence
52	Qwen3.8-Omni-Flash	90.6%	estimated ± 2.7 pp, high confidence
53	DeepSeek V4.1 Flash	90.4%	estimated ± 2.7 pp, high confidence
54	DeepSeek V4 Pro 0813	90.0%	measured
55	Kimi K2.6	89.9%	estimated ± 2.7 pp, high confidence
56	Beam	89.9%	estimated ± 2.7 pp, high confidence
57	Qwen3.7 Plus	89.6%	estimated ± 2.7 pp, high confidence
58	GPT-5.3 Codex	89.1%	estimated ± 5.4 pp, medium confidence
59	DeepSeek V4 Flash 0731	89.0%	measured
60	Interfaze Beta	88.9%	estimated ± 2.7 pp, high confidence
61	GPT-5.6 Luna	88.0%	measured
62	Claude Opus 4.6	87.9%	estimated ± 2.7 pp, high confidence
63	Ornith-1.5-35B-A3B	87.9%	estimated ± 2.7 pp, high confidence
64	Qwen3.8-27B	87.9%	estimated ± 2.7 pp, high confidence
65	Solar Pro 4	87.5%	estimated ± 2.7 pp, high confidence
66	Claude Opus 4.7	87.4%	estimated ± 5.4 pp, medium confidence
67	Grok 4.6	87.0%	measured
68	GPT-6 Luna	86.7%	measured
69	Grok 4.20	86.7%	estimated ± 1.9 pp, high confidence
70	GPT-5.2	86.2%	measured
71	Agents-A1	86.1%	estimated ± 11.9 pp, low confidence
72	Claude Sonnet 4.6	86.0%	measured
73	Inkling	85.7%	estimated ± 2.7 pp, medium confidence
74	Grok 4.5	85.7%	measured
75	MiMo-V2.5-Pro	85.3%	estimated ± 3.7 pp, high confidence
76	Kimi K2.5	85.2%	estimated ± 2.7 pp, medium confidence
77	MiniMax M3	84.9%	estimated ± 5.4 pp, medium confidence
78	Hy3 Preview	84.6%	estimated ± 2.7 pp, medium confidence
79	MiniMax M2.7	84.2%	estimated ± 2.7 pp, medium confidence
80	Nemotron 3 Ultra	84.2%	estimated ± 2.7 pp, medium confidence
81	Inkling-Small	84.0%	measured
82	MiMo-V2-Pro	83.9%	estimated ± 5.4 pp, medium confidence
83	GPT-5.2-Codex	83.6%	estimated ± 5.4 pp, medium confidence
84	Qwen 3.6 Max (preview)	83.4%	estimated ± 5.4 pp, medium confidence
85	Gemini 3 Pro Deep Think	83.3%	estimated ± 1.9 pp, high confidence
86	Ornith-1.5-9B	83.2%	estimated ± 2.7 pp, medium confidence
87	Solar Open 2	83.0%	estimated ± 2.7 pp, medium confidence
88	GLM-5.1	82.8%	estimated ± 2.7 pp, medium confidence
89	GLM-5	82.4%	estimated ± 2.7 pp, medium confidence
90	Muse Spark	82.1%	estimated ± 1.9 pp, high confidence
91	Ternary Bonsai 2 27B	82.0%	estimated ± 2.7 pp, medium confidence
92	A.X K2	81.7%	estimated ± 2.7 pp, medium confidence
93	Ling 3.0 Flash	80.5%	estimated ± 2.7 pp, medium confidence
94	Qwen3.6 Plus	80.4%	estimated ± 5.4 pp, medium confidence
95	Claude Opus 4.5 Thinking	80.0%	measured
96	Quasar 438B	79.7%	estimated ± 5.4 pp, medium confidence
97	GLM-5-Turbo	79.3%	estimated ± 5.4 pp, medium confidence
98	MAI-Thinking-1	79.1%	estimated ± 2.7 pp, medium confidence
99	Apodex 1.1	78.8%	estimated ± 5.4 pp, medium confidence
100	Apodex 1.1 Mini	78.8%	estimated ± 5.4 pp, medium confidence
101	Ling 3.0 Flash FP8	78.7%	estimated ± 2.7 pp, medium confidence
102	Kimi K2.7 Code	77.1%	estimated ± 5.4 pp, medium confidence
103	Hy3	75.5%	estimated ± 5.4 pp, medium confidence
104	K-EXAONE 2.0	75.1%	estimated ± 2.7 pp, medium confidence
105	Qwen3.5 Flash	75.1%	estimated ± 5.9 pp, medium confidence
106	Gemini 3 Pro	75.0%	measured
107	Ling 3.0 Flash VL	72.9%	estimated ± 5.4 pp, medium confidence
108	GPT-5.1	72.8%	measured
109	MiMo-V2.5	72.6%	estimated ± 5.9 pp, medium confidence
110	Gemini 3.1 Flash-Lite	71.5%	estimated ± 5.9 pp, medium confidence
111	MiMo-V2-Omni	70.4%	estimated ± 5.4 pp, medium confidence
112	GPT-5.1-Codex	69.4%	estimated ± 5.4 pp, medium confidence
113	GPT-5.1-Codex-Max	69.4%	estimated ± 5.4 pp, medium confidence
114	Claude Opus 4.5	69.4%	estimated ± 5.4 pp, medium confidence
115	GLM-5V-Turbo	68.6%	estimated ± 5.4 pp, medium confidence
116	Kimi K2.5 (Reasoning)	68.4%	estimated ± 5.4 pp, medium confidence
117	Mercury 2.5	68.2%	estimated ± 2.7 pp, medium confidence
118	Gemma 4 12B	67.8%	estimated ± 2.7 pp, medium confidence
119	GPT-5 (high)	66.2%	estimated ± 5.4 pp, medium confidence
120	Qwen3.5-27B	65.8%	estimated ± 5.4 pp, medium confidence
121	GPT-5 (medium)	65.7%	estimated ± 5.4 pp, medium confidence
122	Claude 4.1 Opus Thinking	65.6%	estimated ± 5.4 pp, medium confidence
123	Command A+	63.9%	estimated ± 5.4 pp, medium confidence
124	Grok 4	63.8%	estimated ± 5.4 pp, medium confidence
125	GPT-5.4 mini	63.7%	measured
126	Claude Sonnet 4.5 Thinking	63.7%	measured
127	GLM-4.7	62.6%	estimated ± 5.4 pp, medium confidence
128	Gemini 3.5 Flash-Lite	62.2%	estimated ± 5.4 pp, medium confidence
129	Trinity-Large-Thinking	62.0%	estimated ± 2.7 pp, medium confidence
130	Claude Sonnet 4.5	61.4%	estimated ± 1.9 pp, high confidence
131	o3-pro	60.6%	estimated ± 5.4 pp, medium confidence
132	Nemotron 3.5 Lightning 30B A3B NVFP4	60.3%	estimated ± 2.7 pp, medium confidence
133	Qwen3.5 397B	58.3%	estimated ± 5.4 pp, medium confidence
134	Qwen3.5 397B (Reasoning)	58.3%	estimated ± 5.4 pp, medium confidence
135	Qwen3.6-27B	58.2%	estimated ± 5.4 pp, medium confidence
136	Nemotron 3 Nano Omni 30B A3B	52.1%	estimated ± 2.7 pp, medium confidence
137	Grok 4.1 Fast (Reasoning)	51.9%	estimated ± 5.4 pp, low confidence
138	GPT-5.4 nano	51.5%	measured
139	o3	50.8%	estimated ± 5.4 pp, low confidence
140	Claude Haiku 4.5	50.8%	estimated ± 5.9 pp, low confidence
141	GLM-4.5	50.8%	estimated ± 5.9 pp, low confidence
142	ZAYA1-8B	49.2%	estimated ± 2.7 pp, medium confidence
143	MiniCPM5-2B	47.2%	estimated ± 2.7 pp, medium confidence
144	Step 3.7 Flash	46.2%	estimated ± 5.4 pp, low confidence
145	LongCat-Flash-Lite-Sparse	45.5%	estimated ± 2.7 pp, medium confidence
146	Qwen3.5-35B-A3B	45.2%	estimated ± 5.4 pp, low confidence
147	Claude 4.1 Opus	40.2%	estimated ± 5.4 pp, low confidence
148	Qwen3 235B 2507	39.5%	estimated ± 6.9 pp, low confidence
149	Qwen3.6-35B-A3B	37.9%	estimated ± 5.4 pp, low confidence
150	Grok 4 Fast (Reasoning)	36.0%	estimated ± 5.4 pp, low confidence
151	Gemini 3 Flash	35.9%	estimated ± 5.4 pp, low confidence
152	Muse Glimmer 30B	32.9%	estimated ± 5.4 pp, low confidence
153	Trinity-Large-Preview	31.0%	estimated ± 2.7 pp, medium confidence
154	Gemma 4 31B	28.5%	estimated ± 3.7 pp, medium confidence
155	Claude 4 Sonnet	27.3%	estimated ± 5.4 pp, low confidence
156	Gemini 2.5 Pro	24.1%	estimated ± 5.4 pp, low confidence
157	MiMo-V2-Flash	23.9%	estimated ± 5.4 pp, low confidence
158	DeepSeek V3.2	23.8%	estimated ± 5.4 pp, low confidence
159	Qwen3 Max	21.3%	estimated ± 5.4 pp, low confidence
160	Qwen3.5-122B-A10B	21.1%	estimated ± 5.4 pp, low confidence
161	Mellum2-12B-A2.5B-Thinking	19.8%	estimated ± 2.7 pp, medium confidence
162	ZAYA1-74B-Preview	19.3%	estimated ± 2.7 pp, medium confidence
163	o1	19.2%	estimated ± 5.4 pp, low confidence
164	GLM-4.6	17.6%	estimated ± 5.4 pp, low confidence
165	Granite 4.2 30B	17.1%	estimated ± 5.4 pp, low confidence
166	Laguna XS.2	15.2%	estimated ± 5.9 pp, low confidence
167	K-Exaone	14.9%	estimated ± 5.4 pp, low confidence
168	Gemma 4 26B A4B	14.8%	estimated ± 3.7 pp, medium confidence
169	Mistral Medium 3.5 128B	14.1%	estimated ± 5.4 pp, low confidence
170	Grok Code Fast 1	13.5%	estimated ± 5.4 pp, low confidence
171	Ling 2.6 Flash	13.4%	estimated ± 5.4 pp, low confidence
172	DeepSeek V3.1	12.0%	estimated ± 5.4 pp, low confidence
173	DeepSeek V3.1 (Reasoning)	11.1%	estimated ± 5.4 pp, low confidence
174	DeepSeek-R1	9.7%	estimated ± 5.4 pp, low confidence
175	Nemotron 3 Super 100B	8.7%	estimated ± 5.4 pp, low confidence
176	Kimi K2	8.4%	estimated ± 5.4 pp, low confidence
177	GPT-4.1	8.3%	estimated ± 5.4 pp, low confidence
178	o3-mini	7.6%	estimated ± 5.4 pp, low confidence
179	o1-pro	7.5%	estimated ± 5.4 pp, low confidence
180	GPT-OSS 120B	5.3%	estimated ± 5.4 pp, low confidence
181	o1-preview	4.8%	estimated ± 5.4 pp, low confidence
182	LLaDA2.2-mini	4.8%	estimated ± 2.7 pp, medium confidence
183	Grok 4.1 Fast	4.6%	estimated ± 5.4 pp, low confidence
184	Mistral Small 4	4.6%	estimated ± 5.4 pp, low confidence
185	Mistral Small 4 (Reasoning)	4.6%	estimated ± 5.4 pp, low confidence
186	GLM-4.5-Air	4.2%	estimated ± 5.4 pp, low confidence
187	Granite 4.2 8B	4.2%	estimated ± 5.4 pp, low confidence
188	Soofi S 30B-A3B	4.2%	estimated ± 2.7 pp, medium confidence
189	Ling 3.0 Tiny	4.2%	estimated ± 5.4 pp, low confidence
190	Mellum2-12B-A2.5B-Instruct	3.0%	estimated ± 2.7 pp, medium confidence
191	GPT-4.1 mini	2.7%	estimated ± 5.4 pp, low confidence
192	Llama 4 Maverick	2.5%	estimated ± 5.4 pp, low confidence
193	North Mini Code	2.4%	estimated ± 5.4 pp, low confidence
194	Gemini 2.5 Flash	2.3%	estimated ± 5.4 pp, low confidence
195	DeepSeek V3 0324	2.1%	estimated ± 5.4 pp, low confidence
196	Mistral Large 3	1.7%	estimated ± 5.4 pp, low confidence
197	Granite 4.2 3B	1.5%	estimated ± 5.4 pp, low confidence
198	Mistral Medium 3	1.5%	estimated ± 5.4 pp, low confidence
199	GPT-OSS 20B	1.4%	estimated ± 5.4 pp, low confidence
200	Gemma 4 E4B	1.3%	estimated ± 5.4 pp, low confidence
201	Nemotron 3 Nano 30B	1.3%	estimated ± 5.4 pp, low confidence
202	Sarvam 105B	1.3%	estimated ± 5.4 pp, low confidence
203	Claude 3 Opus	1.2%	estimated ± 5.4 pp, low confidence
204	DeepSeek V3	1.0%	estimated ± 5.4 pp, low confidence
205	GPT-4o	1.0%	estimated ± 5.4 pp, low confidence
206	LFM2.5-2.6B	1.0%	estimated ± 5.4 pp, low confidence
207	DeepSeek R1 Distill Qwen 32B	1.0%	estimated ± 5.4 pp, low confidence
208	Llama 4 Scout	0.8%	estimated ± 5.4 pp, low confidence
209	Gemini 1.5 Pro	0.7%	estimated ± 5.4 pp, low confidence
210	GPT-4.1 nano	0.7%	estimated ± 5.4 pp, low confidence
211	Solar Pro 3	0.7%	estimated ± 5.4 pp, low confidence
212	Gemma 4 E2B	0.7%	estimated ± 5.4 pp, low confidence
213	Qwen3-Omni-30B-A3B-Thinking	0.6%	estimated ± 5.4 pp, low confidence
214	Ultravox v0.6 Llama 3.3 70B	0.6%	estimated ± 5.4 pp, low confidence
215	Mistral Large 2	0.6%	estimated ± 5.4 pp, low confidence
216	Nemotron Ultra 253B	0.6%	estimated ± 5.4 pp, low confidence
217	Llama 3.1 405B	0.5%	estimated ± 5.4 pp, low confidence
218	LFM2.5-8B-A1B	0.4%	estimated ± 5.4 pp, low confidence
219	GPT-4 Turbo	0.4%	estimated ± 5.4 pp, low confidence
220	Solar Pro 2	0.4%	estimated ± 5.4 pp, low confidence
221	Nova Pro	0.4%	estimated ± 5.4 pp, low confidence
222	Qwen2.5 Coder 32B Instruct	0.3%	estimated ± 5.4 pp, low confidence
223	GPT-4o mini	0.3%	estimated ± 5.4 pp, low confidence
224	Sarvam 30B	0.3%	estimated ± 5.4 pp, low confidence
225	Laguna M.1	0.2%	estimated ± 5.9 pp, low confidence
226	Celeris-1	0.2%	estimated ± 5.4 pp, low confidence
227	Exaone 4.0 32B	0.2%	estimated ± 5.4 pp, low confidence
228	MiniCPM5-1B	0.2%	estimated ± 2.7 pp, medium confidence
229	LFM2.5-230M	0.2%	estimated ± 2.7 pp, medium confidence
230	Qwen3-Omni-30B-A3B-Instruct	0.2%	estimated ± 5.4 pp, low confidence
231	Phi-4	0.2%	estimated ± 5.4 pp, low confidence
232	Phi-4 Multimodal Instruct	0.1%	estimated ± 5.4 pp, low confidence
233	Claude 3 Haiku	0.1%	estimated ± 5.4 pp, low confidence
234	Gemini 1.0 Pro	0.1%	estimated ± 5.4 pp, low confidence
235	Exaone 4.0 1.2B	0.1%	estimated ± 5.4 pp, low confidence
236	Granite-4.0-H-1B	0.1%	estimated ± 5.4 pp, low confidence
237	Gemma 3 27B	0.1%	estimated ± 5.4 pp, low confidence
238	Granite-4.0-350M	0.1%	estimated ± 5.4 pp, low confidence
239	Granite-4.0-H-350M	0.1%	estimated ± 5.4 pp, low confidence
240	LFM2.5-VL-1.6B-Extract	0.1%	estimated ± 5.4 pp, low confidence
241	Claude 3.5 Sonnet	0.0%	estimated ± 6.9 pp, low confidence
242	LFM2.5-VL-450M	0.0%	estimated ± 6.9 pp, low confidence

ARC-AGI-1 leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General