Hacker News Reader

has anyone done a comparison of gpt-4 vs. the gpt-4 that microsoft says is being included in most of it's products these days?

I want to say they both aren't exactly the same, as I noticed through some responses to same questions by each, but I am interested in a detailed analysis if anyone has done that.

View on HN

It's a finetuned version of GPT4 for some scenarios (more chatty, with "Sydney" personality).

They are testing GPT4-Turbo, which seems to be closer (or the same) to ChatGPT style and quality of answer. But there will be a new, separated option to enable "Sydney".

Worth to point out that Balanced is not running pure GPT4. It runs multiple models. Only Creative/Precise runs GPT4 only.

> I want to say they both aren't exactly the same, as I noticed through some responses to same questions by each, but I am interested in a detailed analysis if anyone has done that.

That is very well possible and can be due to one of the many parameters of the LLM being different, like the temperature for instance.

Reply on news.ycombinator.com