choose model by argument

[go-dockly]

It would be useful to have a -m (--model 'gpt-4') argument for sgpt

https://platform.openai.com/docs/models/gpt-4

[TheR1D]

Currently GPT-4 is not available thru API, when OpenAI will allow to use it via API, it will be implemented.

[go-dockly]

is there a possibility to specify a model by cmd arg?
could this be added?

[TheR1D]

Currently sgpt uses /chat API endpoint, which has only two available model options (except coming GPT-4): gpt-3.5-turbo - used by sgpt and gpt-3.5-turbo-0301. I think there is no reason to use gpt-3.5-turbo-0301 with sgpt. Imho it doens't make sense to have --model option for now, but once GPT-4 will be available over API, it will be implemented.

[go-dockly]

coming gpt 4 it has this option gpt-4-32k (Same capabilities as the base gpt-4 mode but with 4x the context length)
so eventually it would be nice to differentiate between those but you are absolutely right it doesn't make sense to differentiate between -0301 versions.

[mdonaberger]

Just commenting to say that I got into the GPT-4 on API preview — I can imagine that it's rolling out now to a lot of people soon.

[lpurdy01]

I have just realized this does have the follow on issue of chats don't keep track of what model they were conducted with. I will have to see how this interacts.

[lpurdy01]

Nvm. It seems to work fine switching models with the same chat_id