The consolidated subscription pitch includes access to 400+ AI models. That’s a lot. But I’m wondering about the practical reality. We have workflows that are optimized for specific models. One workflow uses GPT-4 because it’s really good at understanding context in customer emails. Another uses Claude because it’s better at code generation. A third uses a smaller model because cost and speed matter more than capability.
If we’re paying one flat subscription for all of them, does that mean we can swap models freely? Or are we paying a premium if we’re routing a lot of traffic to the heavyweight models? And what happens when you’re using a model that’s not quite right for the job, just because it’s included in your subscription?
I’m also wondering about coverage gaps. There are specialty models for specific domains—medical language processing, financial analysis, that kind of thing. Does a platform that includes 400+ models actually include the niche models you specifically need, or does it package a bunch of general-purpose models with a few specialty options? And if a model becomes deprecated or gets worse, how do you know if you should switch, or if the platform will deprecate it anyway?
From a cost perspective, does unified pricing actually encourage better model selection, or does it encourage people to just use whatever model is available instead of optimizing for the right tool? And how do you handle the situation where none of the included models are actually ideal for what you’re trying to do?
Has anyone actually used a unified subscription service and hit a situation where the available models weren’t a good fit? How did you handle it?