If we find out in 2024, was o1's Transformer base trained on 10+x as much compute as GPT-4's?
10
Ṁ1kṀ1.1kresolved Jul 2
Resolved
N/A1H
6H
1D
1W
1M
ALL
This question is managed and resolved by Manifold.
Market context
Get 
1,000 to start trading!
1,000People are also trading
Related questions
How fast will you be able to train a GPT-2-level AI on a consumer GPU in 2030?
609
In yottaFLOPs (10^24), how much compute will GPT-4 be trained with?
22
How much compute will be used to train GPT-5?
Will $10,000 worth of AI hardware be able to train a GPT-3 equivalent model in under 1 hour, by EOY 2027?
16% chance
Before 2028, will anyone train a GPT-4-level model in a minute?
29% chance
Will growth in the maximum MW used to train AIs, slow down by more than x2 after GPT-5-like?
55% chance
GPT-4 performance and compute efficiency from a simple architecture before 2026
19% chance
Will growth in the maximum MW used to train AIs, slow down by more than x2 after GPT-6-like?
60% chance
Will a GPT-4 quality model be trained for under $10.000 by 2030?
90% chance
Will OpenAI's autonomous agent be based on GPT-4?
34% chance