chatgpt-api: ChatGPT API stopped working with error message "That model does not exist"

Verify latest release

  • I verified that the issue exists in the latest chatgpt release (confirmed on 4.1.3)

Verify webapp is working

  • I verify that the ChatGPT webapp is working properly for this account.

Environment details

[simon@simon ~]$ node --version 
v19.6.0
[simon@simon Music]$ hostnamectl
 Static hostname: simon
       Icon name: computer-laptop
         Chassis: laptop 💻
         Kernel: Linux 5.15.91-4-lts
    Architecture: x86-64
 Hardware Vendor: Dell Inc.
  Hardware Model: XPS 15 7590
Firmware Version: 1.7.0

Describe the Bug

Earlier today, this library still worked, now even the basic example gives the following error:

/usr/bin/node ./src/test-api.js
Uncaught ChatGPTError Error: ChatGPT error 404: {
    "error": {
        "message": "That model does not exist",
        "type": "invalid_request_error",
        "param": null,
        "code": null
    }
}
    at <anonymous> (/home/simon/git/Misc-Projects/consequences.app/node_modules/chatgpt/build/index.js:202:29)
    at processTicksAndRejections (internal/process/task_queues:95:5)
    --- await ---
    at processTicksAndRejections (internal/process/task_queues:95:5)
    --- await ---
    at runMainESM (internal/modules/run_main:53:21)
    at executeUserEntryPoint (internal/modules/run_main:79:5)
    at <anonymous> (internal/main/run_main_module:23:47)
Process exited with code 1

This runs the following code:

import { ChatGPTAPI } from 'chatgpt'

const api = new ChatGPTAPI({
  apiKey: process.env.OPENAI
})

const res = await api.sendMessage('Hello World!')
console.log(res.text)

It worked earlier today. Might something have changed with the API?

About this issue

  • Original URL
  • State: closed
  • Created a year ago
  • Reactions: 2
  • Comments: 36 (14 by maintainers)

Most upvoted comments

If we finally couldn’t find an available and free chat model, can we turn back to the browser based solution?

We’re working on a workaround.

In the meantime, you can switch back to using text-davinci-003 via https://github.com/transitive-bullshit/chatgpt-api/releases/tag/v4.2.0

OpenAI seems to be trolling us. text-chat-davinci-002-20221122 works again

@qwertyuu the text-davinci-002-render model is unfortunately not available to use from OpenAI’s API. We were using a previously leaked model that’s fine-tuned for chat and was free, but OpenAI just disabled it.

@SimonBaars 100% agreed with your takeaways. Thanks 🙏

We’re working on a workaround. In the meantime, you can switch back to using text-davinci-003 via v4.2.0 (release)

Thank you for the blazingly fast response 😃

I promise I’m not a bot 😂 hahahha

yes, sadly openai shut down the model

If you’re seeing this error, upgrade to latest – but be aware that it’s using text-davinci-003 which will cost money to use.

This is a temporary workaround.

When using text-chat-davinci-002-20221122, it always returned 503 error: That model is currently overloaded with other requests. I remember that the response of this model was relatively fast before. Did OpenAI reduce the computing resources that this model can call?

I saw text-davinci-002-render in the API calls of chat.openai.com

image

I’m just catching up on the linked thread https://github.com/transitive-bullshit/chatgpt-api/issues/292. Some takeaways for newcomers like me:

  • text-davinci-003 is slightly less optimized for the ChatGPT usecase
  • The text-davinci-003 model does incur API costs (!)

We’re working on a workaround.

In the meantime, you can switch back to using text-davinci-003 via https://github.com/transitive-bullshit/chatgpt-api/releases/tag/v4.2.0

Thank you for the blazingly fast response 😃

found my answer image

@nick4fake we provide multiple options and I’ve pushed 80+ releases over the past two months to try and help people use ChatGPT until the official API is released – at which point we’ll replace the underlying impl with the real version.

TL;DR you probably want to use ChatGPTUnofficialProxyAPI which uses the real ChatGPT.

CleanShot 2023-02-19 at 02 25 54@2x-opt

So it’s not a solution 😛 it’s just the old crappy model 🙈

Currently the more stable is the text-davinci-003 model used in version 4.2.0

Yes, but the tiny detail is that it is a paid model.

I agree. It looks like it would be better to continue the work based on 3.5.2