Model and Accessing GPT4Turbo

Hi, I have an gpt4 access and cant get access to gpt4turbo.

When I listmodals, gpt4 is type “openai” and gpt-4-turbo is type “system”.

I got an email, telling me about these great new modals and I cant access them. What does a person need to do, in order to access these modals. Customer Service is TERRIBLE. Document is TERRIBLE.

Why am I being emailed about features that I cant access…

Hey, welcome to the dev community!

What shows for you in the playground?

How are you listing the models?

Sounds the same as everyone

{‘id’: ‘gpt-4-0125-preview’, ‘created’: 1706037612, ‘object’: ‘model’, ‘owned_by’: ‘system’},
{‘id’: ‘gpt-4-turbo-preview’, ‘created’: 1706037777, ‘object’: ‘model’, ‘owned_by’: ‘system’},

You can use either of those two names, exactly.

listing via the API

27

Usable

gpt-4-0613 openai

gpt-4 openai

gpt-3.5-turbo openai

gpt-3.5-turbo-16k-0613 openai

gpt-3.5-turbo-0613 openai

gpt-3.5-turbo-0301 openai

UnUsable

gpt-3.5-turbo-instruct system

dall-e-3 system

dall-e-2 system

gpt-3.5-turbo-0125 system

text-embedding-ada-002 openai-internal

tts-1-hd-1106 system

text-embedding-3-small system

tts-1-hd system

gpt-4-0125-preview system

gpt-4-turbo-preview system

text-embedding-3-large system

whisper-1 openai-internal

babbage-002 system

davinci-002 system

gpt-3.5-turbo-16k openai-internal

gpt-4-1106-preview system

gpt-3.5-turbo-1106 system

gpt-4-vision-preview system

tts-1-1106 system

tts-1 openai-internal

gpt-3.5-turbo-instruct-0914 system

if I use any model owned by ‘system’ it rejects

How about:

{‘id’: ‘gpt-3.5-turbo-0125’, ‘created’: 1706048358, ‘object’: ‘model’, ‘owned_by’: ‘system’}
{‘id’: ‘gpt-4-1106-preview’, ‘created’: 1698957206, ‘object’: ‘model’, ‘owned_by’: ‘system’}

Are you making direct calls using the latest version of the openai library that knows about the newest models?

Using something like langchain that has models hardcoded and old?

The exact error message and error code received?

1 Like

In playground I can access Turbo, but I cant change the ResponseFormat field (not supplied).

I’m using my own library in swift, it returns a generic 400 error code.

There should be a SIMPLE way of determining if you have access to an endpoint.

Frustrating waste of 2 days !

None of those either (in API) all in Playground

What Tier of billing are you? Brand new account?

…vs a slightly seasoned account…

Ones a New Account (hoping that would save the day becuase it said, all accounts after some date) paid $10 Credit.

The other ones 18 months old.

Both have (had) Chat GPT Plus subs.

Under Limits it lists the gpt-4-turbo limits… Currently I’m limited ot ZERO !

“Down for my account or just my code?”

Try local code in a python environment 3.8-3.11, just have your key as environment variable OPENAI_API_KEY:

import os
import requests

headers = {
  "Content-Type": "application/json",
  "Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}"
  # "Authorization": "Bearer sk-yourkeyhere"
}

params = {
  "max_tokens": 10, "top_p": 0.01,
  "messages": [
    {"role": "system","content":
      'One word only. JSON format as response. json key: "word"'},
    {"role": "user","content": "say hello. No markdown."}]
}

for model in ["gpt-4-0314", "gpt-4-0613","gpt-4-1106-preview",
              "gpt-4-0125-preview", "gpt-4-turbo-preview",
              "gpt-4-vision-preview", "gpt-4-badname"]:
    params.update({"model": model})
    response = requests.post("https://api.openai.com/v1/chat/completions",
                             headers=headers,
                             json=params)

    if response.status_code != 200:
        print(f"{model} error {response.status_code}: {response.text}")
    else:
        # print(response.json())
        print(model + response.json()['choices'][0]['message']['content'])

Response:

gpt-4-0314{“word”: “hello”}
gpt-4-0613{“word”: “Hello”}
gpt-4-1106-preview{“word”:“hello”}
gpt-4-0125-preview{“word”:“hello”}
gpt-4-turbo-preview{“word”:“hello”}
gpt-4-vision-preview{“word”: “hello”}
gpt-4-badname error 404: {
“error”: {
“message”: “The model gpt-4-badname does not exist”,
“type”: “invalid_request_error”,
“param”: null,
“code”: “model_not_found”
}
}

Then based on compliance, you can decide if you need JSON mode anyway…

Account problems? Send to “help” in platform.openai.com.

Help tells you to use a generic endpoint gpt-4-preview… which doesnt exist. (404 error).

Please remove and update.

The help widget that pops up has “messages” at the bottom, where you can click through a tree of options, and send a message directly to OpenAI account staff about your account’s model issue if the API call code I demonstrate is failing. (knowledgebase not going to have this issue’s answer)

It seems unlikely python code employing no other libraries would fail if all models are okay in API Playground. You can also generate and employ a new API key, to ensure it has entitlements to all current models.