I sure hope so. Got beta access early on, with a pretty strong use case, namely generating human-readable summaries of records in a large database from meta data. Once I saw the pricing I thought that must be a joke. If you’re just fooling around and create a handful of prompts, okay, but it’s prohibitively expensive for anything starting from a few thousand requests, not even to mention a 100+ million :)

I think we might actually just try GPT-J…