OpenAI o3 at a glance
On the last day of its 12 Days event, OpenAI previewed o3 and o3-mini and opened early access for safety researchers. ARC Prize verified 75.7% on ARC-AGI at its cost limit and 87.5% with far more compute.
What changed with OpenAI o3
On December 20, 2024, the final day of its 12 Days of OpenAI event, OpenAI previewed o3 and o3-mini, the successors to o1. The models were not released publicly. Instead, OpenAI opened applications for safety and security researchers to test them, with applications closing January 10, 2025, and said o3 access would take several more weeks.
The same day, the ARC Prize Foundation reported that a tested o3 version scored 75.7% on the ARC-AGI semi-private evaluation within its $10,000 compute limit, and 87.5% using about 172 times more compute, at a cost of thousands of dollars per task. ARC later noted that the o3 OpenAI released in 2025 differed from the version it tested.
OpenAI o3 in numbers
Values are copied from the linked source on the date shown. They are not IA Compa measurements.
- ARC-AGI semi-private score (within cost limit)as of · Source for ARC-AGI semi-private score (within cost limit): arcprize.org (opens in a new tab)
- 75.7%
Where to compare OpenAI o3
Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.
Leaderboards
General leaderboards
Official resources
Why OpenAI o3 matters for businesses
For businesses, the o3 preview showed how quickly reasoning models were improving, only three months after o1-preview. It also showed that top scores could depend on very expensive compute, so benchmark results needed to be read together with cost.
Sources for this story
- Early access for safety testing (opens in a new tab)Official
OpenAI · openai.com
- Deliberative alignment (opens in a new tab)Official
OpenAI · openai.com
- Early access for safety testing (archived December 21, 2024) (opens in a new tab)Official
OpenAI via Internet Archive · web.archive.org
- OpenAI o3 Breakthrough High Score on ARC-AGI-Pub (opens in a new tab)Benchmark
ARC Prize · arcprize.org
Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.
- #llm
- #openai
- #reasoning
- #o3
- #arc-agi
