@RemNi would a differently named model that clearly acts as GPT4's successor count?

@Bayesian Depends. If they stopped doing the GPT-X series for whatever reason (and told us about it) and started releasing LLMs under another name, then my preference would be to resolve this question either as N/A or NO.

If they started releasing multimodal models under a different branding, but didn't say anything about GPT-5, then I would resolve this question as NO once it went past the deadline.

@Bayesian If it's very clearly GPT-4's successor in all but name, but they say they're going with a different branding from then on (without OpenAI clearly stating that "this is GPT-5"), I wouldn't expect to resolve this question as YES. It would be more of a N/A vs NO dilemma

@Bayesian but there are other potential edge cases that would have to be considered in the future, for instance if they release a model called GPT5-pro or something

@RemNi I see. Altman has said the GPT-N names are kind of lame, and that he'd like to give the future versions a better name, I think? seemed to imply anthropomorphic names were better, and had Sora (more anthropomorphic than say dall-e or chatgpt)

I still expect them to call it the new step after gpt3 and gpt4 or something. and to be multimodal, and for them to say outloud "it's basically gpt5, but with this name"

@Bayesian Yeah it really depends. If they say "this is GPT-5", but give it a different branding name, and everyone including OpenAI refers to the two names interchangeably, then sure, that could be a YES.

@Bayesian But I think if there was significant contention then I'd just go for a N/A