
Before 2028, will any AI model achieve the same or greater benchmarks as o3 high with <= 1 million tokens per question?
7
1kṀ20022028
87%
chance
1D
1W
1M
ALL
Specifically, the key benchmarks here are ARC, Codeforces elo, and Frontier Math score. The relevant scores are 2727 codeforces elo, 87.5% on arc semi-private, and 25.2% on Frontier Math.
The model must achieve these benchmarks while using no more than 1,000,000 reasoning tokens per question on average.
For context, o3 used 5.7B tokens per task to achieve its ARC score. It also scored 75.7% on low compute mode using 33M tokens per task.
https://arcprize.org/blog/oai-o3-pub-breakthrough
Also note that if the final version of o3 has improved or worsened benchmarks the goalposts will not change. The model must beat the benchmarks listed here.
This question is managed and resolved by Manifold.
Get
1,000 to start trading!
Related questions
Related questions
Will any AI get a score of at least 45% on Humanity’s Last Exam benchmark before March 11, 2025?
7% chance
Will there be a significant AI safety incident involving OpenAI o1 before April 2025?
3% chance
Will AI image generating models score >= 90% on Winoground by June 1, 2025?
82% chance
Will any AI model score >80% on Epoch's Frontier Math Benchmark in 2025?
17% chance
Will the top model by OpenAI rank 3rd (or lower) behind 2 other model families at any point before 2026?
41% chance
Will AI top level capabilities generally be judged by question and answer benchmarks in 2029?
25% chance
Will a Chinese-made AI beat o3's December score on Frontier Math by the end of 2025?
73% chance
Will an AI model outperform 95% of Manifold users on accuracy before 2026?
58% chance
Will OpenAI models achieve ≥90% on SimpleBench by the end of 2025?
42% chance
Will any model get above human level on the Simple Bench benchmark before September 1st, 2025.
63% chance