Skip to content

Benchmark

ARC-AGI 2

Scores are percentages. Higher is better.

Models measured
98
Scores recorded
318
Sources
3
IQ weight
1.29

Score distribution

42.4%
Mean
42.5%
Median
35%
Std dev
0.9%
10th percentile
90%
90th percentile
0%
Min
0%95.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.29 for chat models, measured over 98 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1GPT-6.2 Astramax· Epoch AI Benchmarking Hub data95.8%
2GPT-6.2 Astraxhigh· Epoch AI Benchmarking Hub data95.8%
3GPT-6 Astramax· Epoch AI Benchmarking Hub data95%
4GPT-6 AstraOpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.95%
5GPT-6.2 Astrahigh· Epoch AI Benchmarking Hub data94.6%
6Gemini 4 Carbonmedium· Epoch AI Benchmarking Hub data94.2%
7GPT-6.1 Solmax· Epoch AI Benchmarking Hub data94.2%
8Claude Fable 5.2xhigh· Epoch AI Benchmarking Hub data93.8%
9Claude Opus 5.5high· Epoch AI Benchmarking Hub data93.3%
10GPT-6 Astraxhigh· Epoch AI Benchmarking Hub data93.3%
11GPT-6.2 Astramedium· Epoch AI Benchmarking Hub data93.3%
12Claude Fable 5.2max· Epoch AI Benchmarking Hub data92.5%
13Claude Opus 5.5xhigh· Epoch AI Benchmarking Hub data92.5%
14GPT-5.6 Solmax· Epoch AI Benchmarking Hub data92.5%
15GPT-5.6 SolOpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.5%
16GPT-5.6 SolOpus 5 System Card capability summary · adaptive thinking · max effort unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.5%
17GPT-5.6 SolSystem Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.5%
18Gemini 4 Carbonhigh· Epoch AI Benchmarking Hub data92.1%
19GPT-6 Astrahigh· Epoch AI Benchmarking Hub data92.1%
20GPT-6 Astramedium· Epoch AI Benchmarking Hub data92.1%
21Claude Opus 5.5max· Epoch AI Benchmarking Hub data91.7%
22GPT-6.1 Solhigh· Epoch AI Benchmarking Hub data91.7%
23GPT-6.1 Solxhigh· Epoch AI Benchmarking Hub data91.7%
24GPT-6.2 Astralow· Epoch AI Benchmarking Hub data91.7%
25Gemini 4 Argonhigh· Epoch AI Benchmarking Hub data91.3%
26Claude Opus 5max· Epoch AI Benchmarking Hub data90.4%
27Claude Opus 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.4%
28Claude Fable 5.2high· Epoch AI Benchmarking Hub data90.4%
29Claude Opus 5OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.4%
30Claude Opus 5Opus 5 System Card capability summary · adaptive thinking · max effort unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.4%
31Claude Fable 5.1max· Epoch AI Benchmarking Hub data90%
32Claude Fable 5.1OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90%
33Claude Fable 5.1System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90%
34Claude Fable 5.1xhigh· Epoch AI Benchmarking Hub data90%
35GPT-5.6 Solxhigh· Epoch AI Benchmarking Hub data90%
36Claude Fable 5.2medium· Epoch AI Benchmarking Hub data89.6%
37GPT-6 Solmax· Epoch AI Benchmarking Hub data89.6%
38Claude Fable 5OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.2%
39Claude Fable 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.2%
40Claude Fable 5max· Epoch AI Benchmarking Hub data89.2%
41Gemini 3.8 Flashhigh· Epoch AI Benchmarking Hub data89.2%
42Claude Fable 5.1high· Epoch AI Benchmarking Hub data88.8%
43Claude Fable 5xhigh· Epoch AI Benchmarking Hub data88.3%
44Claude Opus 5high· Epoch AI Benchmarking Hub data88.3%
45Claude Fable 5high· Epoch AI Benchmarking Hub data87.5%
46Claude Opus 5.5medium· Epoch AI Benchmarking Hub data87.5%
47Gemini 4 Argonmedium· Epoch AI Benchmarking Hub data87.1%
48Claude Fable 5.2low· Epoch AI Benchmarking Hub data86.7%
49GPT-6.1 Solmedium· Epoch AI Benchmarking Hub data86.7%
50Claude Fable 5.1medium· Epoch AI Benchmarking Hub data86.3%

Showing 1–50 of 318 results.

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · Data by LLM Stats