GPT-3.5 vs. GPT-4: Where is a comparison of features?

I’m interested in a point-by-point comparison of GPT-3.5 vs. GPT-4 and I haven’t been able to find one on OpenAI’s website. Lots of blog posts from third parties, but I surprised I can’t find an official comparison from OpenAI?

For instance, I recently saw someone post that they used ChatGPT to convert a table within a PDF file into a .csv. I wanted to try that out, but GPT-3.5 tells me it cannot accept PDFs.

So, is this a feature that is unlocked with GPT-4? Or is it some plugin?

OpenAI did post on their blog about image analysis features: “ChatGPT can now see, hear, and speak”, but it’s unclear to me if this is exclusive to GPT-4 or even active on GPT-4.

I don’t mind paying for a subscription, but I would like to know what I’m getting for my money. Anyone know of a concise comparison of GPT-3.5 vs… GPT-4?

The pdf ability is not in general release yet, s small subset of users have it to trial, the only real solid difference between 3.5 and 4 is the context size and model inference performance, it’s hard to put any blow blow points, it’s just a larger and more powerful model so it’s able to do more complex tasks and follow more instructions. OpenAi DevDay is on Monday, so I’d watch that as there will be a lot of new announcements to talk about after that.

1 Like