// BENCHMARK

ARC-AGI-1 benchmark

AI model leaderboard for the ARC-AGI-1 benchmark. Compare how large language models score on ARC-AGI-1, see the full ranking, and understand what this AI benchmark measures. GPT-5 currently leads with 65.7. Original Abstraction and Reasoning Corpus — François Chollet's test of fluid, sample-efficient abstraction.

Leaderboard

# Model Organization Score Variant Source
#1 GPT-5 OpenAI 65.7 thinking-high-cited-ring-1t official ↗
#2 Ring-1T Ant Group 55.94 thinking official ↗
#3 Qwen3-235B-A22B-Thinking (Jul 2025) Qwen 48.12 thinking-cited-ring-1t official ↗
#4 gemini-2.5-pro Google DeepMind 45.44 cited-ring-1t official ↗
#5 Ling-1T Ant Group 43.81 pass@1 official ↗
#6 DeepSeek-V3.1-Terminus DeepSeek 40.62 thinking-cited-ring-1t official ↗
#7 Kimi K2 Moonshot 22.19 instruct-0905-cited-ling-1t official ↗
#8 gemini-2.5-pro Google DeepMind 18.94 lowthink-cited-ling-1t official ↗
#9 DeepSeek-V3.1-Terminus DeepSeek 14.69 non-thinking-cited-ling-1t official ↗
#10 GPT-5 OpenAI 14.06 main-cited-ling-1t official ↗

Frequently asked questions about ARC-AGI-1

What is the ARC-AGI-1 benchmark?

Original Abstraction and Reasoning Corpus — François Chollet's test of fluid, sample-efficient abstraction.

How is the ARC-AGI-1 benchmark scored?

ARC-AGI-1 is scored using the accuracy metric, where a higher score is better. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on ARC-AGI-1?

As of the latest reported scores on GenAIList, GPT-5 achieves the highest result on ARC-AGI-1 with a score of 65.7.

Is a higher ARC-AGI-1 score better?

Yes. On ARC-AGI-1 a higher score indicates better performance, so models near the top of the leaderboard are the strongest.