benchgap
Knowledge & reasoning

ARC-AGI-1 leaderboard

As of 2026-11-21, the highest measured score on ARC-AGI-1 is 98.5% by Claude Fable 5. 244 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Muse Wonder Full99.4%estimated ± 7.8 pp, low confidence
2Claude Fable 598.5%measured
3Gemini 3.8 Flash98.5%measured
4Gemini 4 Argon98.5%measured
5GPT-6.2 Astra98.5%measured
6Claude Fable 5.297.8%estimated ± 3.4 pp, high confidence
7Claude Fable 5.197.5%measured
8Claude Opus 597.5%measured
9Claude Opus 5.597.5%measured
10Gemini 4 Carbon97.5%measured
11GPT-6.2 Sol97.5%measured
12GPT-6 Astra97.5%measured
13Claude Mythos 596.9%estimated ± 3.9 pp, high confidence
14Claude Sonnet 5.596.7%estimated ± 3.9 pp, high confidence
15Sakana Fugu96.7%estimated ± 2.5 pp, high confidence
16Sakana Fugu-Ultra96.7%estimated ± 2.5 pp, high confidence
17GPT-5.6 Sol96.5%measured
18GPT-5.6 Terra96.5%measured
19GPT-6.1 Sol96.5%measured
20Inkling 2 Small96.5%measured
21Gemini 4 Neon96.5%estimated ± 7.9 pp, low confidence
22Muse Spark 1.196.3%estimated ± 3.9 pp, high confidence
23Qwen4 Max96.2%estimated ± 3.9 pp, high confidence
24Muse Wonder95.9%estimated ± 7.8 pp, low confidence
25Pareto 26.995.8%estimated ± 3.9 pp, high confidence
26Claude Sonnet 595.7%estimated ± 0.8 pp, low confidence
27Hy495.5%estimated ± 11.6 pp, low confidence
28Gemini 3.7 Flash95.5%measured
29GPT-6 Sol95.5%measured
30ERNIE 5.5 Flash Preview95.2%estimated ± 3.9 pp, high confidence
31dots3-note Preview95.2%estimated ± 1.7 pp, high confidence
32Kimi K3.1 Pro95.1%estimated ± 2.5 pp, high confidence
33GPT-5.595.0%measured
34GPT-5.5 Pro95.0%measured
35MiMo-V2.6-Pro94.7%estimated ± 7.8 pp, low confidence
36GPT-5.4 Pro94.5%measured
37Kimi K394.5%measured
38Step 5 Preview94.2%estimated ± 2.5 pp, high confidence
39MiniMax M3.1 Flash94.2%estimated ± 3.9 pp, high confidence
40Gemini 3.1 Pro94.1%estimated ± 1.7 pp, high confidence
41Claude Opus 4.793.8%estimated ± 1.7 pp, high confidence
42GPT-5.493.7%measured
43GPT-6.1 Terra93.4%estimated ± 4.4 pp, low confidence
44Muse Spark 1.393.4%estimated ± 7.8 pp, low confidence
45Ornith-1.5-397B93.3%estimated ± 2.5 pp, high confidence
46Qwen3.8 Max93.0%estimated ± 2.5 pp, high confidence
47Claude Opus 4.6 (Adaptive)93.0%measured
48Pareto 26.10 Preview92.8%estimated ± 2.5 pp, high confidence
49Qwen3.7 Max92.8%estimated ± 2.5 pp, high confidence
50Hy4 flash92.7%estimated ± 11.6 pp, low confidence
51Hy4 preview92.6%estimated ± 2.5 pp, high confidence
52Claude Opus 4.892.5%measured
53Gemini 3.5 Flash92.5%measured
54Qwen3.8-Flash-Next91.8%estimated ± 2.5 pp, high confidence
55Gemini 4 Hydrogen91.4%estimated ± 7.9 pp, low confidence
56Gemini 3.6 Flash91.2%measured
57GLM-5.291.1%estimated ± 2.5 pp, high confidence
58MiniMax M390.9%estimated ± 6.5 pp, medium confidence
59Qwen3.8-Omni-Flash90.8%estimated ± 2.5 pp, high confidence
60DeepSeek V4.1 Flash 101790.6%estimated ± 2.5 pp, high confidence
61Beam 1.0 501B-A23B90.2%estimated ± 2.5 pp, high confidence
62Kimi K2.690.1%estimated ± 2.5 pp, high confidence
63DeepSeek V4 Pro90.0%measured
64Particle-189.9%estimated ± 2.5 pp, high confidence
65Qwen3.7 Plus89.8%estimated ± 2.5 pp, high confidence
66Beam 1.0 501B-A23B FP889.8%estimated ± 2.5 pp, high confidence
67Sage-A1.589.7%estimated ± 7.8 pp, low confidence
68Grok 4.789.5%measured
69Interfaze Beta89.1%estimated ± 2.5 pp, high confidence
70Nemotron 3.5 Super VL 120B A12B NVFP489.1%estimated ± 2.5 pp, high confidence
71Grok 4.389.0%estimated ± 6.5 pp, medium confidence
72Claude Haiku 5.589.0%measured
73DeepSeek V4 Flash 073189.0%measured
74Nemotron 3.5 Super VL 120B A12B88.6%estimated ± 2.5 pp, high confidence
75Ling 3.1 Flash88.2%estimated ± 7.8 pp, low confidence
76Beam 1.0 501B-A23B NVFP488.2%estimated ± 2.5 pp, high confidence
77Mercury 388.1%estimated ± 2.5 pp, high confidence
78Claude Opus 4.688.1%estimated ± 2.5 pp, high confidence
79Ornith-1.5-35B-A3B88.1%estimated ± 2.5 pp, high confidence
80Qwen3.8-27B88.1%estimated ± 2.5 pp, high confidence
81GPT-5.6 Luna88.0%measured
82Muse Spark 1.288.0%estimated ± 7.8 pp, low confidence
83Qwen3.8 Max Preview88.0%estimated ± 7.8 pp, low confidence
84Solar Pro 487.7%estimated ± 2.5 pp, high confidence
85Nemotron 3.5 Bolt 74B-A9B BF1687.5%estimated ± 2.5 pp, high confidence
86GPT-5.3 Codex87.4%estimated ± 7.8 pp, low confidence
87Grok 4.687.0%measured
88Grok 4.2086.7%estimated ± 1.7 pp, high confidence
89GPT-6 Luna86.7%measured
90OrcaCyber-Nano-1.086.6%estimated ± 2.5 pp, high confidence
91Agents-A186.4%estimated ± 11.6 pp, low confidence
92GPT-5.286.2%measured
93Claude Sonnet 4.686.0%measured
94Inkling86.0%estimated ± 2.5 pp, medium confidence
95Muse Spark Open85.9%estimated ± 7.8 pp, low confidence
96Grok 4.585.7%measured
97Kimi K2.585.5%estimated ± 2.5 pp, medium confidence
98MiMo-V2.5-Pro85.1%estimated ± 3.9 pp, high confidence
99Hy3 Preview84.8%estimated ± 2.5 pp, medium confidence
100MiniMax M2.784.4%estimated ± 2.5 pp, medium confidence
101Nemotron 3 Ultra84.4%estimated ± 2.5 pp, medium confidence
102Inkling-Small84.0%measured
103Inkling 2 Mini83.7%estimated ± 2.5 pp, medium confidence
104MiMo-V2.6-Flash83.7%estimated ± 7.8 pp, low confidence
105GLM-5.383.7%estimated ± 6.5 pp, medium confidence
106Grok 483.7%estimated ± 6.5 pp, medium confidence
107Nemotron 3.5 Bolt 74B-A9B83.4%estimated ± 2.5 pp, medium confidence
108Ornith-1.5-9B83.4%estimated ± 2.5 pp, medium confidence
109Gemini 3 Flash83.4%estimated ± 6.5 pp, medium confidence
110Solar Open 283.2%estimated ± 2.5 pp, medium confidence
111GLM-5.183.0%estimated ± 2.5 pp, medium confidence
112GLM-582.7%estimated ± 2.5 pp, medium confidence
113Qwen3.6 Plus82.5%estimated ± 6.5 pp, medium confidence
114Ternary Bonsai 2 27B82.2%estimated ± 2.5 pp, medium confidence
115Kimi K2.7 Code82.2%estimated ± 7.8 pp, low confidence
116Mistral Large 482.2%estimated ± 7.8 pp, low confidence
117Muse Spark82.0%estimated ± 1.7 pp, high confidence
118A.X K281.9%estimated ± 2.5 pp, medium confidence
119Quasar 438B81.8%estimated ± 7.8 pp, low confidence
120GPT-5.2-Codex81.3%estimated ± 7.8 pp, low confidence
121Ripple-65B Preview81.2%estimated ± 2.5 pp, medium confidence
122Ling 3.0 Flash80.8%estimated ± 2.5 pp, medium confidence
123GLM-5.3-Flash80.7%estimated ± 6.5 pp, medium confidence
124Claude Opus 4.5 Thinking80.0%measured
125o3-pro79.8%estimated ± 7.9 pp, low confidence
126MAI-Thinking-179.3%estimated ± 2.5 pp, medium confidence
127GPT-579.2%estimated ± 6.5 pp, medium confidence
128GPT-5.1-Codex79.0%estimated ± 7.8 pp, low confidence
129GPT-5.1-Codex-Max79.0%estimated ± 7.8 pp, low confidence
130Ling 3.0 Flash FP878.9%estimated ± 2.5 pp, medium confidence
131Grok 4 Fast (Reasoning)78.8%estimated ± 6.5 pp, medium confidence
132Hy378.4%estimated ± 7.8 pp, low confidence
133Apodex 1.178.2%estimated ± 7.8 pp, low confidence
134Apodex 1.1 Mini78.2%estimated ± 7.8 pp, low confidence
135Claude 4.1 Opus78.0%estimated ± 7.9 pp, low confidence
136Qwen3.5 397B77.6%estimated ± 6.7 pp, medium confidence
137Qwen 3.6 Max (preview)77.5%estimated ± 7.8 pp, low confidence
138Muse Glimmer 30B76.7%estimated ± 7.8 pp, low confidence
139Grok 4.1 Fast (Reasoning)76.7%estimated ± 6.5 pp, medium confidence
140Step 3.7 Flash76.5%estimated ± 7.8 pp, low confidence
141o376.3%estimated ± 6.5 pp, medium confidence
142Ling 3.0 Flash VL76.2%estimated ± 7.8 pp, low confidence
143Gemini 2.5 Flash75.8%estimated ± 7.8 pp, low confidence
144Qwen3.6-27B75.7%estimated ± 6.7 pp, medium confidence
145Gemini 3.5 Flash-Lite75.7%estimated ± 6.5 pp, medium confidence
146Claude 4 Sonnet75.6%estimated ± 7.8 pp, low confidence
147K-Exaone75.6%estimated ± 7.8 pp, low confidence
148MiMo-V2-Omni75.6%estimated ± 7.8 pp, low confidence
149DeepSeek V3.275.4%estimated ± 7.8 pp, low confidence
150Trinity-Large-Preview75.4%estimated ± 7.8 pp, low confidence
151Trinity-Large-Thinking75.4%estimated ± 7.8 pp, low confidence
152K-EXAONE 2.075.4%estimated ± 2.5 pp, medium confidence
153Gemma 4 E4B75.2%estimated ± 7.8 pp, low confidence
154GLM-5V-Turbo75.2%estimated ± 7.8 pp, low confidence
155Gemini 3 Pro75.0%measured
156Command A+75.0%estimated ± 7.8 pp, low confidence
157GLM-5-Turbo75.0%estimated ± 7.8 pp, low confidence
158Granite 4.2 30B75.0%estimated ± 7.8 pp, low confidence
159Granite 4.2 8B75.0%estimated ± 7.8 pp, low confidence
160MiMo-V2-Pro75.0%estimated ± 7.8 pp, low confidence
161MiniCPM5-2B75.0%estimated ± 7.8 pp, low confidence
162Mistral Small 475.0%estimated ± 7.8 pp, low confidence
163North Mini Code75.0%estimated ± 7.8 pp, low confidence
164Sarvam 30B75.0%estimated ± 7.8 pp, low confidence
165Celeris-174.7%estimated ± 7.8 pp, low confidence
166Claude 3.5 Haiku74.7%estimated ± 7.8 pp, low confidence
167Claude 3 Haiku74.7%estimated ± 7.8 pp, low confidence
168DeepSeek V374.7%estimated ± 7.8 pp, low confidence
169DeepSeek V3 032474.7%estimated ± 7.8 pp, low confidence
170Devstral Small74.7%estimated ± 7.8 pp, low confidence
171Exaone 4.0 1.2B74.7%estimated ± 7.8 pp, low confidence
172Exaone 4.0 32B74.7%estimated ± 7.8 pp, low confidence
173Gemini 2.5 Flash-Lite74.7%estimated ± 7.8 pp, low confidence
174Gemma 3 27B74.7%estimated ± 7.8 pp, low confidence
175Gemma 4 E2B74.7%estimated ± 7.8 pp, low confidence
176GLM-4.5-Air74.7%estimated ± 7.8 pp, low confidence
177GPT-4.1 nano74.7%estimated ± 7.8 pp, low confidence
178GPT-4o74.7%estimated ± 7.8 pp, low confidence
179Granite-4.0-350M74.7%estimated ± 7.8 pp, low confidence
180Granite-4.0-H-1B74.7%estimated ± 7.8 pp, low confidence
181Granite-4.0-H-350M74.7%estimated ± 7.8 pp, low confidence
182Granite 4.2 3B74.7%estimated ± 7.8 pp, low confidence
183Grok 4.1 Fast74.7%estimated ± 7.8 pp, low confidence
184Grok Code Fast 174.7%estimated ± 7.8 pp, low confidence
185LFM2.5-2.6B74.7%estimated ± 7.8 pp, low confidence
186LFM2.5-8B-A1B74.7%estimated ± 7.8 pp, low confidence
187LFM2.5-VL-1.6B-Extract74.7%estimated ± 7.8 pp, low confidence
188Ling 2.6 Flash74.7%estimated ± 7.8 pp, low confidence
189Ling 3.0 Tiny74.7%estimated ± 7.8 pp, low confidence
190Llama 3.1 405B74.7%estimated ± 7.8 pp, low confidence
191Llama 3.1 70B74.7%estimated ± 7.8 pp, low confidence
192Llama 3.3 70B74.7%estimated ± 7.8 pp, low confidence
193Llama 4 Maverick74.7%estimated ± 7.8 pp, low confidence
194Llama 4 Scout74.7%estimated ± 7.8 pp, low confidence
195Magistral Small74.7%estimated ± 7.8 pp, low confidence
196Mercury 2.574.7%estimated ± 7.8 pp, low confidence
197Mistral Large 274.7%estimated ± 7.8 pp, low confidence
198Mistral Medium 374.7%estimated ± 7.8 pp, low confidence
199Mistral Medium 3.5 128B74.7%estimated ± 7.8 pp, low confidence
200Mistral Small 3.174.7%estimated ± 7.8 pp, low confidence
201Nova Pro74.7%estimated ± 7.8 pp, low confidence
202Phi-474.7%estimated ± 7.8 pp, low confidence
203Phi-4-reasoning74.7%estimated ± 7.8 pp, low confidence
204Qwen3.5-4B74.7%estimated ± 7.8 pp, low confidence
205Qwen3-Omni-30B-A3B-Instruct74.7%estimated ± 7.8 pp, low confidence
206Qwen3-Omni-30B-A3B-Thinking74.7%estimated ± 7.8 pp, low confidence
207Sarvam 105B74.7%estimated ± 7.8 pp, low confidence
208Solar Pro 274.7%estimated ± 7.8 pp, low confidence
209Solar Pro 374.7%estimated ± 7.8 pp, low confidence
210Ultravox v0.6 Llama 3.3 70B74.7%estimated ± 7.8 pp, low confidence
211o1-preview74.2%estimated ± 7.9 pp, low confidence
212Qwen3.5 Flash73.6%estimated ± 6.5 pp, medium confidence
213o1-mini73.4%estimated ± 7.9 pp, low confidence
214QwQ-32B73.2%estimated ± 7.9 pp, low confidence
215GPT-5.172.8%measured
216Claude 3 Opus72.8%estimated ± 7.9 pp, low confidence
217DeepSeek R1 Distill Qwen 32B72.6%estimated ± 7.9 pp, low confidence
218Gemini 1.5 Pro72.4%estimated ± 7.9 pp, low confidence
219Qwen3.5-122B-A10B72.0%estimated ± 6.7 pp, medium confidence
220Gemini 1.5 Flash71.9%estimated ± 7.9 pp, low confidence
221GPT-4 Turbo71.9%estimated ± 7.9 pp, low confidence
222Qwen2.5 Coder 32B Instruct71.7%estimated ± 7.9 pp, low confidence
223GPT-4o mini71.7%estimated ± 7.9 pp, low confidence
224DeepSeek-V2.571.7%estimated ± 7.9 pp, low confidence
225MiMo-V2.571.1%estimated ± 6.5 pp, medium confidence
226Gemini 1.0 Pro71.0%estimated ± 7.9 pp, low confidence
227Qwen3.6-35B-A3B70.1%estimated ± 6.7 pp, medium confidence
228Gemini 3.1 Flash-Lite70.1%estimated ± 6.5 pp, medium confidence
229Qwen3.5-27B68.5%estimated ± 6.7 pp, medium confidence
230GLM-4.767.9%estimated ± 6.5 pp, medium confidence
231Claude Opus 4.566.5%estimated ± 6.5 pp, medium confidence
232Qwen3 Max66.5%estimated ± 6.5 pp, medium confidence
233Qwen3.5-35B-A3B64.4%estimated ± 6.7 pp, medium confidence
234GPT-5.4 mini63.7%measured
235Claude Sonnet 4.5 Thinking63.7%measured
236Muse Ember63.5%estimated ± 7.9 pp, low confidence
237MiMo-V2-Flash62.9%estimated ± 6.7 pp, medium confidence
238Claude Sonnet 4.561.4%estimated ± 1.7 pp, high confidence
239Phi-4 Multimodal Instruct61.2%estimated ± 8.1 pp, low confidence
240Gemini 2.5 Pro60.7%estimated ± 6.7 pp, medium confidence
241Llama 3.1 8B60.6%estimated ± 8.1 pp, low confidence
242Claude 3.5 Sonnet58.1%estimated ± 8.4 pp, medium confidence
243o3-mini57.3%estimated ± 6.5 pp, low confidence
244Claude 3.7 Sonnet Thinking56.8%estimated ± 6.5 pp, low confidence
245GLM-4.654.9%estimated ± 6.5 pp, low confidence
246o4-mini54.9%estimated ± 6.5 pp, low confidence
247DeepSeek-R1-052854.4%estimated ± 6.7 pp, low confidence
248Grok 3 Mini52.2%estimated ± 6.7 pp, low confidence
249o151.8%estimated ± 6.5 pp, low confidence
250GPT-OSS 120B51.6%estimated ± 6.7 pp, low confidence
251GPT-5.4 nano51.5%measured
252Claude 4 Opus Thinking50.0%estimated ± 6.7 pp, low confidence
253Claude Haiku 4.549.3%estimated ± 6.5 pp, low confidence
254GLM-4.549.3%estimated ± 6.5 pp, low confidence
255Nemotron 3 Super 100B48.9%estimated ± 6.7 pp, low confidence
256o1-pro48.2%estimated ± 6.7 pp, low confidence
257Qwen3 235B 250743.5%estimated ± 6.7 pp, low confidence
258Mistral Large 340.3%estimated ± 6.5 pp, low confidence
259GPT-4.1 mini39.1%estimated ± 6.5 pp, low confidence
260Nemotron Ultra 253B38.8%estimated ± 6.7 pp, low confidence
261Claude 3.7 Sonnet37.9%estimated ± 6.5 pp, low confidence
262Grok 3 [Beta]36.9%estimated ± 6.7 pp, low confidence
263GLM-4.7-Flash36.3%estimated ± 6.7 pp, low confidence
264Kimi K235.9%estimated ± 6.7 pp, low confidence
265Claude 4 Opus35.3%estimated ± 6.7 pp, low confidence
266DeepSeek V3.135.3%estimated ± 6.7 pp, low confidence
267LFM2.5-24B-A2B33.8%estimated ± 6.7 pp, low confidence
268GPT-4.133.4%estimated ± 6.5 pp, low confidence
269Gemini 2.0 Flash33.0%estimated ± 6.5 pp, low confidence
270LFM3-1.7B31.1%estimated ± 7.9 pp, low confidence
271Laguna XS.230.0%estimated ± 8.4 pp, low confidence
272Nemotron 3 Nano 30B29.4%estimated ± 6.7 pp, low confidence
273Laguna M.129.1%estimated ± 8.4 pp, low confidence
274Gemma 4 31B28.0%estimated ± 3.9 pp, medium confidence
275Nemotron 3 Nano Omni 30B A3B26.9%estimated ± 6.7 pp, low confidence
276DeepSeek-R124.7%estimated ± 6.7 pp, low confidence
277GPT-OSS 20B24.7%estimated ± 6.7 pp, low confidence
278Qwen3 235B-A22B (Reasoning)23.4%estimated ± 6.7 pp, low confidence
279ZAYA1-8B23.1%estimated ± 6.7 pp, low confidence
280Spark-X2.6-4B22.9%estimated ± 3.9 pp, medium confidence
281Magistral Medium22.5%estimated ± 6.7 pp, low confidence
282GPT-4.518.4%estimated ± 6.7 pp, low confidence
283Phi-4-reasoning-plus16.5%estimated ± 6.7 pp, low confidence
284Nemotron 3.5 Lightning 30B A3B NVFP414.0%estimated ± 3.9 pp, medium confidence
285Gemma 4 26B A4B13.7%estimated ± 3.9 pp, medium confidence
286Gemma 4 12B13.6%estimated ± 3.9 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General