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.
1,000People are also trading
@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?
@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