Is ChatGPT more up to date than GPT3.5?

API
1

When I do a side by side comparison of data:
Screen Shot 2023-01-12 at 5.43.12 PM

and

Screen Shot 2023-01-12 at 5.43.29 PM
Screen Shot 2023-01-12 at 5.43.29 PM1018Ă—232 13.9 KB

I get two separate answers. Does ChatGPT contain knowledge that GPT3.5 does not? Is the version of davinci in the playground not 3.5?

2

We’re not entirely sure about ChatGPT, but we think it’s possibly GPT-3.5.

The latest model in Playground is TD3 (Text-Davinci-003), which is also “kinda-sorta” a GPT-3.5 version?

Me and others are fairly certain they’re different models, though.

Hope this helps.

3 Likes
3

I was personally under the impression that text-embedding-ada-002 was the only available model built on GPT-3.5

Edit: Shows what I know!

1 Like
4

So you don’t think any of the completion endpoints are 3.5?

Edit: wrong phrase

5

I mean, maybe TD3 (text-davinci-003) is GPT-3.5 (although unofficially…)

The model naming convention is evolving… Things are moving so quickly now…

ETA: I defer to the post below with a source! Thanks.So hard to keep up these days!

2 Likes
6

“GPT 3.5” is what they’ve started calling their models that combine code and text completion.

text-davinci-003, text-davinci-002 and code-davinci-002 are all considered “GPT 3.5”

https://beta.openai.com/docs/model-index-for-researchers/models-referred-to-as-gpt-3-5

3 Likes
7

This is perfect! Thanks all!