Attempting to use gpt-3.5-turbo, getting gpt-3.5-turbo-0301 instead

General API discussion
1

Using the Python API like this:

 chat = openai.ChatCompletion.create(
                model="gpt-3.5-turbo",
                messages=messages,
                temperature=.9,
               )

This works, but in the response it is reporting using model gpt-3.5-turbo-0301.
In the docs is states that 0301 does not pay much attention to the system message, and in fact I have had a bit of trouble with it ignoring parts of my system message.

Why is it not using the non-0301 version like I asked for? Thanks.

1 Like
2

Hi @tmh,

comparing your code with the Documentation from Microsoft and one Example on Git Hub there is one thing you can try to modify in your code to fix this issue:

  • Change “chat” → “response”

Sources where “respnse” is used:
chatgpt-3.5-turbo/main.py at main ¡ kydycode/chatgpt-3.5-turbo ¡ GitHub
Quickstart - Using the ChatGPT API - Azure OpenAI Service | Microsoft Learn

3

Thanks for getting back with me linus. Are you talking about the variable name I bind to the response from the ChatCompletion.create method? That name is completely arbitrary, I can’t see how that could change anything. :neutral_face:

1 Like
4

Hi @tmh

My understanding is that gpt-3.5-turbo will always be the latest model, which is gpt-3.5-turbo-0301 as of now. Hence it shows gpt-3.5-turbo-0301 in the response. Once an updated version is released gpt-3.5-turbowill point to it.

image
image1560×542 70.7 KB

5 Likes
5

Ah yes there I was going in the wrong direction! Thanks :slight_smile:

I found an answer to your question:
→ In the request you specify with “gpt-3.5-turbo” that you want to use the latest version of the model. This is currently gpt-3.5-turbo-0301, so this is the model you see in the response. Should there be 0302 in the future, then this would be the model used in the answer.

Refer to this thread:
Model usage says ‘gpt-3.5-turbo-0301’ but API call uses ‘gpt-3.5-turbo’ , any idea why? : OpenAI (reddit.com)

6

I see. That is odd because of that comment in the docs about the 0301 version not paying attention to system messages. Of course, it could be the case that the non-0301 doesn’t either and it just wasn’t explicitly mentioned. Turn out I had a typo in my system message that may have been a significant contributor to the problem because after its correction the issue seems to have resolved. :flushed:
Thank you all for your help.

7

Yes prompt design takes some iterations. Glad you found one that works.

8

that comment in the docs about the 0301 version not paying attention to system messages. Of course, it could be the case that the non-0301 doesn’t either and it just wasn’t explicitly mentioned.

I think you’re misunderstanding the note in the docs. I made the same error. What they’re saying is that the current latest model, which is the 0301 version, has that issue. Subsequent versions of the model do not have that issue, but subsequent versions are also not yet available.