What will the successor to o3 be called?
➕
Plus
57
Ṁ36k
2026
12%
o4
0.4%
o5
83%
GPT-5
0%
o-xp
0%
o-vista
0%
o9
0.1%
agi-1
5%
Other

A successor only counts if they have a name that isn't o3.x or o3-x or something that contains o3. o4 counts, o5 counts, reasoning models with a different name pattern that are clearly better than o3 count.

  • Update 2025-04-17 (PST) (AI summary of creator comment): Benchmark Results as Evidence:

    • The successor does not need to be officially released; showing benchmark results similar to those provided for o3 before January when the market was created is sufficient.

    • This clarification means that a demonstration of performance through benchmarking can count as identifying a successor even without a formal launch.

  • Update 2025-07-17 (PST) (AI summary of creator comment): The successor must be a major model that is widely considered a successor to o3. It will not resolve to simply the next model that is released.

  • Update 2025-07-17 (PST) (AI summary of creator comment): The creator has specified that ChatGPT agent does not count as a successor to o3.

  • Update 2025-08-01 (PST) (AI summary of creator comment): The position of a model in the ChatGPT model selector will not be a significant factor for resolution. The key criterion remains what model is widely considered to be the successor.

Get Ṁ1,000 play money
Sort by:

is this just the next model that is released?

@realDonaldTrump no, the next major model that is widely considered a successor to o3

@Bayesian what is widely considered?

@realDonaldTrump roughly "something people generally think of as"

@Bayesian This is the model selector for ChatGPT Pro Plan

Let's say GPT-5 was added and became the first listed model and 4o moved down to "more models" or somewhere lower down the list but o3 stayed as #2 the whole time

What would this resolve to?

I would support a close when gpt-5 is released and poll for resolution

@realDonaldTrump Depends what it is widely considered to be, I don’t think its position in the model selector is going to affect resolution much though

@Bayesian can it be closed and then a poll?

@realDonaldTrump it could be, but likely won't be

do you think ChatGPT agent counts? It seems like a smarter model overall than o3, and not just at computer use, though it's not a clear Pareto improvement. Unlike o3 it's not available as a chat/general model but that's immaterial per your criteria.

system card here: https://cdn.openai.com/pdf/6bcccca6-3b64-43cb-a66e-4647073142d7/chatgpt_agent_system_card_launch.pdf

@JoshYou Good question, I think it doesn't count

@JoshYou It's not a model. It is a tool used with all other models.

@Bayesian Send statement that says not model

@Bayesian See I can have agent enabled with 4o or o3

@Bayesian

From the legend itself

the models do not reliably know who they are or what tools they are using, so this is not trustworthy

or o3 + search's analysis of the situation
https://chatgpt.com/s/t_688d3540c67c819187ac99c1e47ac844

it's an o3 fine-tune. so a different model, but only somewhat

opened a Ṁ100 GPT-5 YES at 20% order

am I missing something? if GPT-5 comes out before o4, is there any reason it wouldn't be the successor to o3? if it's not a reasoning model or does worse than o3 on benchmarks? (both seem very unlikely)

@JoshYou I guess this is downstream of the fact that I don't believe there will be a a model called o4, because if GPT-5 is a superior reasoning model to o3 then what would o4 be?

Does o4 option also apply to o4-mini?

i don't think o4-mini counts as a successor to o3 since it seems worse on most benchmarks but o4-mini kinda implies the existence of o4 so i am considering resolving early, what do ppl think

bought Ṁ50 GPT-5 YES

@Bayesian I’d say to give it time, since OAI might call GPT-5 their next-gen reasoner in lieu of a new reasoner in the o series

@Bayesian I understood "successor" as "the next release that Open AI tells people to use". So this would only count if o4 is actually released.

But I agree it's ambiguous...

@TimothyJohnson5c16 I would lean toward release of the successor not being required, like if they show benchmark results for o4 in a similar way that they did for o3 right before january when this market was created, that would make it a successor

@TimothyJohnson5c16 I would lean toward release of the successor not being required, like if they show benchmark results for o4 in a similar way that they did for o3 right before january when this market was created, that would make it a successor