What I was meaning was that ChatGPT is not a foundation model since it's been fine-tuned. Although the definition in the link is sufficiently broad you could choose to include it.

I can't tell from the OpenAI docs whether it's possible to access GPT-4 without the ChatGPT fine-tuning. If so, that'd make this result more meaningful. Otherwise, I just don't think you can draw any great conclusions from this.

The instruction fine tuning is what manifests knowledge and reasoning.

GPT is general purpose, it's not fine tuned for specific topics. A fine tuned model is tuned to a specific subject.