I understand why you are not in a hurry but I'd like to offer a counter-argument. I have a machine powerful enough to experiment with local LLMs and interesting

I understand why you are not in a hurry but I'd like to offer a counter-argument. I have a machine powerful enough to experiment with local LLMs and interestingly enough at least some models are on a par with what OpenAI used to charge for (GPT4). It means I can experiment and see what these models are capable of, whether it makes sense to use them to solve my problem, and if not - why, and if anything can be done about it on my part. And because the API is the same, it's relatively easy. It's not necessary to be at the bleeding edge, two steps away is also fine, but today LLMs are just another tool in your toolbelt and I reached the conclusion it makes no sense to ignore them, in spite of how irritating the hype (or, sometimes, plain lies) can be.