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
251GPT-5 nanohigh· Epoch AI Benchmarking Hub data2.6%
252Gemini 2.5 Flash Preview 05-20 (no think)23k· Epoch AI Benchmarking Hub data2.5%
253o4 Minimedium· Epoch AI Benchmarking Hub data2.4%
254Seed 2.0 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim2.3%
255Gemini 2.5 Flash Preview 05-20 (no think)1k· Epoch AI Benchmarking Hub data2.2%
256Claude Sonnet 48k· Epoch AI Benchmarking Hub data2.1%
257Gemini 2.5 Flash Preview 05-20 (no think)8k· Epoch AI Benchmarking Hub data2.1%
258DeepSeek V4 Flash 0731none· Epoch AI Benchmarking Hub data2.1%
259DeepSeek V4.1 Flash 1017none· Epoch AI Benchmarking Hub data2.1%
260o3 Minimedium· Epoch AI Benchmarking Hub data2.1%
261O3 Prolow· Epoch AI Benchmarking Hub data2%
262OpenAI o3low· Epoch AI Benchmarking Hub data2%
263Gemini 2.5 Flash Preview 05-20 (no think)16k· Epoch AI Benchmarking Hub data2%
264GPT 5low· Epoch AI Benchmarking Hub data1.9%
265GPT 5.1low· Epoch AI Benchmarking Hub data1.9%
266GPT 5.4 Nanomedium· Epoch AI Benchmarking Hub data1.9%
267O3 Promedium· Epoch AI Benchmarking Hub data1.9%
268Gemini 2.5 Flash Preview 05-20 (no think)· Epoch AI Benchmarking Hub data1.7%
269Claude Haiku 4.58k· Epoch AI Benchmarking Hub data1.7%
270GPT-5 miniminimal· Epoch AI Benchmarking Hub data1.7%
271o4 Minilow· Epoch AI Benchmarking Hub data1.7%
272GPT-6 Solnone· Epoch AI Benchmarking Hub data1.7%
273GPT 5.4 Nanolow· Epoch AI Benchmarking Hub data1.5%
274Gemini 3.5 Flash Litelow· Epoch AI Benchmarking Hub data1.5%
275Qwen 3.8 27Bnone· Epoch AI Benchmarking Hub data1.5%
276DeepSeek R1· Epoch AI Benchmarking Hub data1.3%
277Gemini 2.0 Flash· Epoch AI Benchmarking Hub data1.3%
278Claude Opus 4· Epoch AI Benchmarking Hub data1.3%
279Claude Sonnet 4· Epoch AI Benchmarking Hub data1.3%
280Codex Mini· Epoch AI Benchmarking Hub data1.3%
281Claude Haiku 4.5· Epoch AI Benchmarking Hub data1.3%
282Claude Haiku 4.51k· Epoch AI Benchmarking Hub data1.3%
283DeepSeek V4.1 Flashnone· Epoch AI Benchmarking Hub data1.3%
284Qwen3 235B A22B Instruct 2507· Epoch AI Benchmarking Hub data1.3%
285GPT-5.4 minilow· Epoch AI Benchmarking Hub data1.1%
286Claude Sonnet 3.78k· Epoch AI Benchmarking Hub data0.9%
287GPT-5 nanomedium· Epoch AI Benchmarking Hub data0.88%
288Claude Sonnet 41k· Epoch AI Benchmarking Hub data0.85%
289DeepSeek V4 Pro 0813none· Epoch AI Benchmarking Hub data0.833%
290Gemini 3.5 Flash Liteminimal· Epoch AI Benchmarking Hub data0.833%
291GPT 5.2· Epoch AI Benchmarking Hub data0.83%
292GPT-5 minilow· Epoch AI Benchmarking Hub data0.83%
293O1 Mini· Epoch AI Benchmarking Hub data0.83%
294gemini-1.5-pro-002· Epoch AI Benchmarking Hub data0.8%
295GPT-4.5 Preview· Epoch AI Benchmarking Hub data0.8%
296Claude Sonnet 3.716k· Epoch AI Benchmarking Hub data0.7%
297GPT 4.1· Epoch AI Benchmarking Hub data0.42%
298GPT 5.1none· Epoch AI Benchmarking Hub data0.42%
299Grok 3 Minilow· Epoch AI Benchmarking Hub data0.42%
300Claude Sonnet 3.71k· Epoch AI Benchmarking Hub data0.4%

Showing 251–300 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