First post measures o3 at high reasoning effort. Second measures at medium reasoning effort. It’s the same model, then and now.

Yeah, I totally get the confusion here. Unfortunately I can't give the recipe behind our models, so there's going to be some irreducible blurriness here, but the following statements are all true:

- o3 pro is based on o3

- o3 pro uses the same underlying model as o3

- o3 pro is similar to o3, but is a distinct thing that's smarter and slower

- o3 pro is not o3 with longer reasoning

In my analogy, o3 pro vs o3 is more than just an input parameter (e.g., not just the accelerator input) but less than a full difference in model (e.g., Ford Mustang vs F150). It's in between, kind of like car trim with the same body but a stronger engine. Imperfect analogy, and I apologize if this doesn't feel like it adds any clarity. At the end of the day, it doesn't really matter how it works - what matters is if people find it worth using.

o3 pro is based on o3 and its style and outputs will be quite similar to o3.

As an analogy, think of it like this:

o3-low ~ Ford Mustang with the accelerator gently pressed

o3-medium ~ Ford Mustang with the accelerator pressed

o3-high ~ Ford Mustang with the accelerator heavily pressed

o3 pro ~ Ford Mustang GT

Even though a Mustang GT is a different car than a Mustang, you don’t give it a totally different name (eg Palomino). The similarity in name signals it has a lot of the same characteristics but a souped up engine. Same for o3 pro.

Fun fact: before GPT-4, we had a unified naming scheme for models that went {modality}-{size}-{version}, which resulted in names like text-davinci-002. We considered launching GPT-4 as something like text-earhart-001, but since everyone was calling it GPT-4 anyway, we abandoned that system to use the name GPT-4 that everyone had already latched onto. Kind of funny how our original unified naming scheme made room for 999 versions, but we didn't make it past 3.

Edit: When I say the Mustang GT is a different car than a Mustang - I mean it literally. If you bought a Mustang GT and someone delivered a Mustang with a different trim, you wouldn't say "great, this is just what I ordered, with the same features/behavior/value." That we call it a different trim is a linguistic choice to signal to consumers that it's very similar, and built on the same production line, but comes with a different engine or different features. Similar to o3 pro.

o3 is a model, and reasoning effort (high/medium/low) is a parameter that goes into the model.

o3 pro is a different thing - it’s not just o3 with maximum remaining effort.

It's the same model, no quantization, no gimmicks.

In the API, we never make silent changes to models, as that would be super annoying to API developers [1]. In ChatGPT, it's a little less clear when we update models because we don't want to bombard regular users with version numbers in the UI, but it's still not totally silent/opaque - we document all model updates in the ChatGPT release notes [2].

[1] chatgpt-4o-latest is an exception; we explicitly update this model pointer without warning.

[2] ChatGPT Release Notes document our updates to gpt-4o and other models: https://help.openai.com/en/articles/6825453-chatgpt-release-...

(I work at OpenAI.)