OpenAI said this about GPT-5.2:

GPT‑5.2 was built in collaboration with our long-standing partners NVIDIA and Microsoft. Azure data centers and NVIDIA GPUs, including H100, H200, and GB200-NVL72, underpin OpenAI’s at-scale training infrastructure, driving significant gains in model intelligence. Together, this collaboration allows us to scale compute with confidence and bring new models to market more quickly.

Of course, GPT-5 was earlier so maybe not GB200

We might not get confirmatiom but the answer will be H100/H200s. Blackwell only started shipping in volume in early 2025.

What is the result if no model named "GPT-5" is ever released?

@Fay42 if they abandon the GPT-N naming scheme, I will resolve based on the next OpenAI model which is reported to be significantly larger than GPT-4

If OpenAI never releases such a model then this market resolves N/A

@ahalekelly as far as I know this is not an actual chip it's possible to buy. It's a description of a theoretical chip of every component of the h100 worked but that's not a thing Nvidia will sell in the real world. Same for other G-series.

@Fay42 oh hmm yeah I forgot that Nvidia referred to the dies using the G prefix, like GH100, GA100, GA102 etc.

What I was referring to here was not that, but what Nvidia calls the “GH200 Grace Hopper Superchip” which is a Grace CPU and Hopper GPU in one package. Very confusing that there’s two overlapping naming schemes. But I do think you’re on to something, I think there are GH200 and GB200 Superchips but I was mistaken and no GH100 Superchip ever existed.