Option to use different GPT models

This is of interest to us mostly because we're all for cost optimizations on the models. When it's our API key, it impacts our costs in the same way, just at scale. The main reason Botsheets works so well for lead generation is because we're using GPT-4. If we give people the ability to just switch to a worse, and cheaper model like 3.5, then you end up getting what you pay for. Here is another chatbot solution we played with that clearly doesn't use GPT-4, or anything comparable. This isn't exactly a good customer experience, and so our users will just end up blaming Botsheets and churning if it's anything like this:

You can compare that UX to what Botsheets delivers now with GPT-4 and it's like comparing a mule to a race horse. So, what would need to happen is not the ability to change the model to GPT-3.5, or just a cheaper model, but a model that's comparable in performance to GPT-4 but just cheaper. The thing is, anything around the core customer experience we have at the moment takes higher priority, and while that does include optimizations for model cost, it's more of an ongoing project behind the scenes rather than a configurable option we extend in the interface. We would need to spend time experimenting with, and validating assumptions with other models, and by the time we get around to completing that R&D, it's possible that GPT-4 costs come down even more to GPT-3.5. Maybe there is GPT-5 at which point giving people the option to switch between the latest and the minimum viable UX (GPT-4) and which would be the cheaper, but acceptable model is possible.

That said, if you have a preference for a specific model that you know of that can be quantified as cheaper and on par with GPT-4, we can take a look.

Thanks for weighing in!