ARC-AGI-3
Abstract reasoning puzzles; scores depend heavily on harness
knowledge and reasoningactive
Definition
- Version
- not in the notes
- Category
- knowledge and reasoning
- Maintainer
- ARC Prize
- Status
- active
- Home page
- https://arcprize.org
Sources
- OpenAI Launches GPT-6 Astra, Pricing, Benchmarks and Who Gets It First pasqualepillitteri.it https://pasqualepillitteri.it/en/news/14246/openai-launches-gpt-6-astra-pricing-benchmarks
- GPT-6 Astra Is Live: OpenAI's New Flagship Open for Calls - API易文档中心 docs.apiyi.com https://docs.apiyi.com/en/news/gpt-6-astra-launch
Scores
| Model | Score | Who ran it | Setting | View or population | Date | Source |
|---|---|---|---|---|---|---|
 GPT-6 Astra | 99.9 % | vendor-reported | with OpenAI adapter harness (apiyi citing Latent Space) | – | – | docs.apiyi.com |
| GPT-6 Astra | 98.6 % | vendor-reported | Conflicting figures across sources | – | – | pasqualepillitteri.it |
| GPT-6 Astra | – % | vendor-reported | direct, about 63 to 66% (apiyi citing Latent Space) | – | – | docs.apiyi.com |
Where this comes from
Scores from different settings, harnesses or dates are not directly comparable; the row says which. Sorted by score, highest first, as a reading aid, not a ranking.
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "ARC-AGI-3", https://aurion.semai.id/benchmarks/arc-agi-3/, records of October 2026.