Please, how can I be sure of which ChatGPT version (4, 4o…) my Custom GPT works on, once I select GPT-4o in the Chat area?
Thanks
The AI model is not yours to pick as creator, but is part of the user’s platform.
It will likely stay at GPT-4 (a gpt-4-turbo obviously switched in), because using gpt-4o would likely break thousands of GPTs at once.
You can see in the network data there is a provision for model in a GPT definition, like one for voice, but it is not used.
{"gizmo":{"id":"g-abcdedf","organization_id":"org-xxx","short_url":"g-xxxx","author":{"user_id":"user-xxxxx","display_name":"URL","link_to":"URL","is_verified":true,"selected_display":"name","will_receive_support_emails":true,"display_socials":[]},"voice":{"id":"ember"},"workspace_id":null,"model":null,"instructions":"# Instructions\n\n## Identity and Purpose\n\...
Thanks both of you for your prompt, yet opposite, responses.
The announcenent states GPT will run on 4o but I’d like any stronger evidence in the page itself.
Any way to assess that?
Thanks both of you!
The OpenAI help site mentions.
GPTs for paid users are currently powered by GPT-4. GPTs for paid users will switch to use GPT-4o in the next several weeks.
3 Likes
At least for me custom GPTs in ChatGPT are using the omni model.
{
"message": {
"id": "1f3ab17f-961e-43dd-8834-f7907b6c3354",
"author": {
"role": "assistant",
"name": null,
"metadata": {}
},
"create_time": 1716422009.504234,
"update_time": null,
"content": {
"content_type": "text",
"parts": [
"Hi there! How can I help you today?"
]
},
"status": "finished_successfully",
"end_turn": true,
"weight": 1,
"metadata": {
"finish_details": {
"type": "stop",
"stop_tokens": [
200002
]
},
"citations": [],
"gizmo_id": "g-RFS6wS6VL",
"is_complete": true,
"message_type": "next",
"model_slug": "gpt-4o",
"default_model_slug": "gpt-4o",
"pad": "AAAAAAAAAAAAAAAAAAAAAAAAA",
"parent_id": "aaa2fdcd-7120-4750-81c6-ab98aa137ff0",
"model_switcher_deny": [
{
"slug": "text-davinci-002-render-sha",
"context": "regenerate",
"reason": "unsupported_gizmo",
"description": "This GPT is not supported"
}
]
},
"recipient": "all"
},
"conversation_id": "8d31ac61-ae26-4e82-bc49-d04af49235f3",
"error": null
}
Source: https://chatgpt.com/share/9e951d61-1b01-40b9-a5e7-a3f448170cd5
1 Like
Excellent help - very clear! Thanks a lot!
1 Like
I’m glad, and I think it’s extraordinary that all GPTs will automatically upgrade to take advantage of the latest model.
This is the right decision, as if you are doing research or deploying mission critical GPTs to team, and retaining a previous version is that important to you, you should be developing your own app in the playground and with the API. (In fact, I’m doing this.)
I like that the consumer facing chat product upgrades automatically. My “personal” GPTs will all benefit from the upgrades… recipebots, helpbots, personabots. They just get more and more capable and knowledgable.
1 Like
That’s a different feature altogether. It has nothing to do with gpt-4o.
Any update on that ? Can someone from open answer this question ?
Also, if you have an “old” customized GPT will/can it be updated to a new model ?
As a Plus subscriber I would like to know, thanks.
1 Like
to the community!