To avoid complications in resolving this market, I define an OpenAI model as announced when it appears on https://openai.com/research under "Models." For reference, here is the current list, as of March 18:
This list looks complete enough to me that I'm comfortable using it as a proxy for model announcements. However, note that e.g. GPT-3.5 would not have counted as a new model, but OpenAI Codex would have.
I may bet in this market.
Update 2025-03-15 (PST) (AI summary of creator comment): Resolution Criteria Update:
The market will now use a common sense interpretation of what constitutes a new model rather than strictly following the appearance under the 'Models' heading on OpenAI’s research page.
All GPT-related entries, including GPT-5, will not be counted as a new model since they are now grouped under the 'Tech' category.
1,000🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ151 | |
| 2 | Ṁ149 | |
| 3 | Ṁ111 | |
| 4 | Ṁ62 | |
| 5 | Ṁ24 |
@traders OpenAI's research index has been reworked. Now all of the GPTs are under the same category, so GPT-5 doesn't count as a "new model."
I thought about resolving blindly based on the research index to stick to the original resolution criteria, but then I noticed that the model names appear under the heading "Tech" rather than "Models," so technically there's no way to fulfill the original criteria anyway.
I think I will just use my own intuitive interpretation of what a model is. In this case, I think the models so far would be GPT-4o, o1, o3, Operator, and GPT-4.5. Does that sound reasonable?
@CDBiddulph what you are saying is intuitive, it would be odd to resolve this market as “1 model release” (o1). Changing resolution criteria so sharply though, and given that you are also trading in this market, makes me think that this should be N/A’d.
@aleven I already sold my position without analyzing whether the new rules would be good for me. I think what most people really care about is the title of the question, "how many AI systems will OpenAI announce in one year." I think N/A might be reasonable, but it's generally discouraged unless you really have to, so I'm not sure.
If you think Alessandro's suggestion of N/A is better, please like his comment, otherwise if you prefer my proposal, please like my top-level comment. If there are no other proposals, I'll respect whatever the majority is, biased towards my own proposal in the event of a tie
I would really expect o1 to count as a new model, but it looks like OpenAI has neglected to update their website... We have a few months, but let me know what you all think I should do if it still isn't updated by then. Or if anyone knows somebody from OpenAI, tell them to get their act together lol
@CDBiddulph o1 is now in the list. Would you still take the list given by "filter" at https://openai.com/research/index/ as the resolution criterion for this market?
@aleven It looks like all GPT models were collapsed into one category... It seems kind of ridiculous to resolve this way, since GPT-4.5 is definitely a new model, and o1 and o3 are also pretty clearly different models. I'll come up with a plan and post in a top-level comment