#GPT-4 API working but not being listed as an available model

12 messages · Page 1 of 1 (latest)

hidden wolf
#

I have a GPT-4 Key from an approved org.

When checking it up with this command:

  -H "Authorization: Bearer $key"```
I get:
```"id": "gpt-4",
   "object": "model",
   "created": 1687882411,
   "owned_by": "openai",
   "permission": [
   {
     "id": "modelperm-Y76uTj6sGxB1dsWvmSDNaUil",
      "object": "model_permission",
      "created": 1687882411,
      "allow_create_engine": false,
      "allow_sampling": false,
      "allow_logprobs": false,
      "allow_search_indices": false,
      "allow_view": false,
      "allow_fine_tuning": false,
      "organization": "*",
      "group": null,
      "is_blocking": false
    }
  ],
  "root": "gpt-4",
  "parent": null```
But when using the key I get results, with no error code/response.

**Am I checking this in the wrong place?**
dusty kernel
#

Have you tried running the same call but against one of the other models that are working like GPT3.5? Take a look at the "allow_view" attribute. Are those true or false? When you look at your rate limits on platform dot do you see GPT-4 listed?

hidden wolf
# 
  "id": "gpt-3.5-turbo",
  "object": "model",
  "created": 1677610602,
  "owned_by": "openai",
  "permission": [
    {
      "id": "modelperm-LeYYAMp5mu0Qkfui5StsNiib",
      "object": "model_permission",
      "created": 1687914041,
      "allow_create_engine": false,
      "allow_sampling": true,
      "allow_logprobs": true,
      "allow_search_indices": false,
      "allow_view": true,
      "allow_fine_tuning": false,
      "organization": "*",
      "group": null,
      "is_blocking": false
    }
  ],
  "root": "gpt-3.5-turbo",
  "parent": null
}
#

gpt-4 is indeed listed on my rate limit page
but isn't it a static page? and doesn't everyone get to have a go at gpt-4 in playgrounds?

dusty kernel
#

I used to have gpt-4 in my rate limit page. I do not now, nor is it visible in Playground to me anymore.

#

I see that the allow_sampling, allow_logprobs, and allow_view are all true for gpt-3.5-turbo but not for gpt-4. That might explain why you can hit the gpt-4 endpoint but not get anything substantive back. Let us know if you see gpt-4 working again. I'm also curious if anyone else has lost connectivity to GPT-4 ether by model-drop or change of permissions.

dusty python
#

you should set org-id in headers

dusty kernel
dusty python
dusty kernel
#

In past, this returned all models I had access to including GPT-4. Did they change requirements to mandate org in header on listmodels to get GPT-4 if I have acces?

dusty python