The counter point to this is that cheaper has rarely been a winning formula in tech. Software/better models cost money, and most use cases are begrudgingly happ

The counter point to this is that cheaper has rarely been a winning formula in tech. Software/better models cost money, and most use cases are begrudgingly happy with “working” software. If a cheaper llm exists, but it isn't as good - then the customer has to spend time evaluating it. Typically, these evaluations don't yield a tangible result.