ARC-AGI-1

accuracy on the first edition of the abstract grid reasoning benchmark

GPT-6 Astra by OpenAI holds the best ARC-AGI-1 result among tracked models, at 98.5% (released Sep 3 2026). Higher scores are better on this benchmark.

1. GPT-6 Astra (OpenAI) — 98.5%; 2. GPT-5.5 (OpenAI) — 95%; 3. GPT-5.4-Pro (OpenAI) — 94.5%; 4. GPT-5.4 (OpenAI) — 93.7%; 5. GPT-5.2 (OpenAI) — 86.2%.

5 tracked models have a published ARC-AGI-1 score. Scores are collected from published lab reports and benchmark sources, with recorded sources credited alongside the results. We do not re-run the evaluations.