jackMort / ChatGPT.nvim

ChatGPT Neovim Plugin: Effortless Natural Language Generation with OpenAI's ChatGPT API
Apache License 2.0
3.67k stars 311 forks source link

Can't get it to work with the GPT-4 model #186

Closed Vanclief closed 1 year ago

Vanclief commented 1 year ago

First of all thanks for the plugin @jackMort, its very cool.

So I got access to the GPT-4 API, so I am trying to enable it in the plugin but I am not being able to. I tried by updating the config:

  openai_params = {
    model = "gpt-4",

  },

And by just editing it directly on the chat settings as stated in here https://github.com/jackMort/ChatGPT.nvim/issues/104#issuecomment-1499509201

But neither option worked for me

image

eyalk11 commented 1 year ago

You take chatgpt word for it? It is probably working

Vanclief commented 1 year ago

Interesting, probably true. I checked in the webapp Chat version and it was returning the correct version. But in the playground it doesn't return the correct version.

Galicarnax commented 1 year ago

Setting model to "gpt-4" in openai_params results in API ERROR: The model: `gpt-4` does not exist (I do have Plus subscription).

jackMort commented 1 year ago

As openai says, it's limited beta, there's wait-list https://openai.com/waitlist/gpt-4-api to join

wstewarttennes commented 1 year ago

@jackMort I have access to the gpt-4 plugin (i think all Plus members do as of today) but still seeing this same issue after adding "gpt-4" to my config. I was getting API ERROR: The model:gpt-4does not exist before today but now i'm getting the same issues as the original poster where the chat bot tells me it's using gpt-3 not 4 (but online it says it's using gpt-4 when i ask it the same question)

Screenshot 2023-07-06 at 4 24 02 PM Screenshot 2023-07-06 at 4 25 06 PM
Vanclief commented 1 year ago

@wstewarttennes I think that its actually gpt-4 even if the response to the prompt its different. Seems that when using the Chat GPT UI there are some extra embedings than when calling it directly from the API or from the playground.

wstewarttennes commented 1 year ago

@Vanclief thanks for the info -- how do you determine if it's using the correct one? the documentation here states it uses 3.5-turbo so i was worried it was hard coded

Ryel commented 1 year ago

@Vanclief thanks for the info -- how do you determine if it's using the correct one? the documentation here states it uses 3.5-turbo so i was worried it was hard coded

You can double check which model you're using from the Daily Usage Breakdown section here: https://platform.openai.com/account/usage

Screenshot 2023-07-13 at 1 34 38 PM