Resolves N/A if OpenAI ceases to exist or they change numbering model
Resolves YES on release of 4.5 model strictly before a release of a model named gpt 5
Resolves NO on release of gpt 5 which isn't preceded by a release of 4.5. Resolves NO in case of a simultaneous release
1,000
3.00The last time Jimmy predicted GPT-4.5 he got it totally wrong, but maybe this time… https://x.com/apples_jimmy/status/1833595024543781088
probably won't be called GPT-4.5 but a major new release before GPT-5 is looking more likely
GPT Next is probably a stand-in for whatever model they release next.
They reportedly want to release a strawberry-powered model this fall, before they release Orion, which is a big new model. Orion will probably be called GPT-5 and i have no idea what they'll call the strawberry model.
OpenAI Dev Day (three dates in October-November) announcement: "While we know developers are waiting for our next big model, which we shared has begun training earlier this year, these events will focus on advancements in the API and our dev tools"
Doesn't strictly preclude a model dropping before or around the Dev Days.
It doesn't preclude it at all. There are a limited amount of spots available, so they don't want anyone unnecessarily taking up a slot because they were hoping for a product announcement there.
This is however yet another (small) sign that they are planning their next model release to be their "next big model" that they announced earlier this year; which is obviously not going to be called GPT-4.5.
"[...] which is obviously not going to be called GPT-4"
Why is that obvious?
seems to me like an odd choice to call your next big model, trained on a huge new (prob >100k H100s) supercomputer the unexciting name GPT-4.5. They‘ve been calling it „GPT-NEXT“ in their slides, contrasted with the GPT-3 and GPT-4 era models, so GPT-5 or something else entirely seems like the next step (they count GPT-3.5 as part of the GPT-3 era)
placing a 500 limit order at 50% if you are interested
@jgyou basically yes, I disbelieve the leaks. The leakers on this topic don't have a great track record, as most leaks made it seem like release was imminent, but it didn't happen.
Those who want to weight leaks highly also have to contend with some of them being about GPT-5 (I don't expect simultaneous release of GPT-4.5 and GPT-5, I think if a GPT-5 release happens it will mean there will be no GPT-4.5), and one of the most prominent leakers saying "GPT-4.5 is cancelled".
The fact that GPT-3.5 exists doesn't IMHO set a strong precedent that 4.5 will. It's early days in their versioning numbers and I don't think they've settled into a rhythm.
It does seem to be true that a release blog post for "GPT-4.5 Turbo" did exist briefly on OpenAIs website last September when search engine crawlers indexed it. But I don't think that is much evidence for anything - that predates GPT-4 Turbo's release, and it had a nonsensical knowledge cutoff in the future - I think it must have been just a test of some aspect of the blog or something like that.
@jgyou short answer: The leaks are not consistent with what you would expect to be leaking if GPT-Whatever would be announced soon. If they're not more likely under that hypothesis, we don't need to update on them.
@chrisjbillington my personal theory is that gpt-4-turbo used to be called gpt-4.5 earlier in the product cycle. Would explain some leaks from before its release.