Will GPT-5 perform better than o1 (not preview) at AIME 2024, Codeforces elo, GPQA, or the 2024 ioi?
Basic
3
Ṁ402025
66%
chance
1D
1W
1M
ALL
o1's scores are the ones mentioned here. https://openai.com/index/learning-to-reason-with-llms/
Resolves yes if OpenAI or a third party tester is able to get GPT-5 to achieve higher scores on any of these benchmarks without any kind of prompt engineering or agent scaffolding
This question is managed and resolved by Manifold.
Get
1,000
and3.00
Sort by:
I had a previous question that asked whether I would judge GPT-5 smarter than 01 that is sitting at 85% yes.
Personally I am unsure why people think this since it is heavily implied GPT-5 won't utilize the scaled test-time compute paradigm which was so key to o1's success. I would bet that GPT-5 will be unable to do this and that the goal is more so achieving near o1 performance more quickly and economically
Related questions
Related questions
Will an AI win a gold medal on the IOI (competitive programming contest) before 2025?
7% chance
Will GPT-5 reach a 1000 rating on Codeforces?
86% chance
Will GPT-5 win Bronze or better at IMO 2025?
24% chance
What will be true about GPT-5?
Will I judge GPT-5 to be smarter than o1 (not preview) after both are released?
77% chance
Will GPT-5 be able to get gold on the International Mathematical Olympiad?
23% chance
Will GPT-5 have a rating of at least 2000 in chess?
55% chance
Will GPT-5 be more competent than me in my area of expertise?
42% chance
Will a 15 billion parameter LLM match or outperform GPT4 in 2024?
24% chance
Will an open-source LLM beat or match GPT-4 by the end of 2024?
82% chance