I have set the “model” parameter to “gpt-3.5-turbo”.
But the program actually calls “gpt-3.5-turbo-0301”.
I want to know how to get the program to call “gpt-3.5-turbo”.
I am in urgent need of help and I beg you to give me a solution.
1 Like
It’s the same thing. It just replies with the latest version. It’s using.
I use the “gpt-3.5-turbo” model, so the returned data should also be “gpt-3.5-turbo” normal ah, why I now return the data is “GPt-3.5-turbo-0301”?
The “3.5-turbo” is the official one, and the “3.5-turbo-0301” is the temporary one, planned to be deprecated in a time to be defined. They have the same update, and the endpoints are (almost) the same - same capabilities as well, they work under a kind of redundancy system.
Some say that “text-davinci-002” model works as an endpoint of “3.5-turbo” (official) with some specific capabilities of that model, maybe embeddings or something. I can’t confirm it, but you can replicate the test and see what you can get.
When you use gpt-3.5-turbo you are telling API to use the latest version of gpt-3.5-turbo
So if you use gpt-3.5-turbo it will use gpt-3.5-turbo-0301
When they update gpt-3.5-turbo-0301, it will use the updated one without you doing anything
If you use gpt-4, it will use the latest gpt-4-0314
If you want to use specific version of the model then you use directly
gpt-3.5-turbo-0301 or gpt-4-0314
There arent really any choices at the moment, only one version available for the actual 3.5 turbo or 4 (although text-davinci-003 text-davinci-002 code-davinci-002 also seems to be under 3.5)
From: OpenAI API
I know that “3.5-turbo-0301” is temporary, so I don’t want to use “3.5-turbo-0301”, I want to use “gpt-3.5-turbo”.
I have set the “model” parameter to “gpt-3.5-turbo”, but my program is still using “3.5-turbo-0301”.
I want to know what I should do to get my program to use “gpt-3.5-turbo”.
“Temporary” is probably not the best word for them to use, but you’re using the latest model. Nothing is wrong or broken. “gpt-3.5-turbo” just points to the most recent 3.5 version, which as of now is 0301. If they release a new one it will point to the new one.
Ok, @NianBroken
Everybody here advised you that is not a problem to use gpt-3.5-turbo-0301. But maybe your professor (shifu), boss (laoban), or customer, is not as brilliant as you: “Oh, my God, you used the 0301 instead turbo - I’ll give you an F- (or Zero) - you’re a bad student” or “I clearly expressed TURBO! I will not pay you… and you’re FIRED!”
It’s possible that the OpenAI API is using multiple instances of the same model to handle the high quantity of requests - a lot of people are using gpt-3.5-turbo directly or through davinci-002 or -003.
They work in redundancy iteration as one of them is busy sending the requests to the other.
Each instance is given a unique identifier. In this case, it looks like there are two instances of the gpt-3.5-turbo model: gpt-3.5-turbo and gpt-3.5-turbo-0301. Both instances should be equivalent in terms of performance, training, endpoints, and capabilities.
You should be able to use either one - and nobody will see the difference in the results… maybe your teacher (shifu), boss (laoban), or customer if they could read your network log directly - your code has the correct model gpt-3.5-turbo.
I agree with @novaphil that “temporary” is not the best word - this is an OpenAI word, but fear not. It means the models are temporary snapshots
| MODEL NAME | DEPRECATION DATE |
|---|---|
| gpt-3.5-turbo-0301 | TBD |
| gpt-4-0314 | TBD |
| gpt-4-32k-0314 | TBD |
The developers can continue using them for at least three months after an updated model has been introduced. Maybe gpt-3.5-turbo will become a gpt-4-free-for-all and gpt-3.5-turbo-0301 will be retired, who knows? And when? Nobody knows (that is what TBD means: To Be Defined).
The most capable and cost-effective model in the GPT-3.5 family is gpt-3.5-turbo which has been optimized for chat but works well for traditional completions tasks as well.
| LATEST MODEL | DESCRIPTION | MAX TOKENS | TRAINING DATA |
|---|---|---|---|
| gpt-3.5-turbo | Most capable GPT-3.5 model and optimized for chat at 1/10th the cost of text-davinci-003. Will be updated with our latest model iteration. | 4,096 tokens | Up to Sep 2021 |
| gpt-3.5-turbo-0301 | Snapshot of gpt-3.5-turbo from March 1st 2023. Unlike gpt-3.5-turbo, this model will not receive updates, and will be deprecated 3 months after a new version is released. | 4,096 tokens | Up to Sep 2021 |
You can see they are twin brothers, but the first will update in the future, and the other will be retired three months after the update of the first. And that is all.
The gpt-3.5-turbo didn’t have any notable updates since March, the only notable events were the exponential increase of users’ requests, degraded performance, and maybe a few outages - meanwhile the gpt-3.5-turbo-0301 is running smoothly because users are fear to use it.
Anyway, if you insist to see gpt-3.5-turbo in your network log - there are tests you can do:
- Check if you have full access to it in your
platformrate limits: Rate limits - OpenAI API
- Check your access to it in this specific API/playground: Playground - OpenAI API - it is a playground for
3.5-turbo and gpt-4 familiesonly.
And select thegpt-3.5-turbo, and make a prompt or two to see if it is working for you:
- Change your code to use the endpoint URL of
gpt-3.5-turbo- assuming the API cannot change the model stated on the URL - usingcurl classes:
... ...
$postData = json_encode($postData);
$curl = curl_init();
curl_setopt_array($curl, array(
CURLOPT_URL => "https://api.openai.com/v1/chat/completions?model=gpt-3.5-turbo",
... ...
Notice this endpoint URL /v1/chat/completions is specific for 3.5-turbo and gpt-4 families only.
Try these tests, and let us know the results. I hope you get what you want.
Also note that in the Playground when you select Chat> Gpt 3.5 and look at network traffic, the request has:
"model":"gpt-3.5-turbo"
and header response has:
openai-model: gpt-3.5-turbo-0301
This is standard behavior.
2 Likes