I mean, a large majority of "pro jev" posts I see purely focus on the speed/cost and take their word on the accuracy. The entire product is "we guarantee the response will fit this schema, and will do it fast and cheap". Personally, I dont see how anything in this class can be "production ready" as a general classifier without the ability to fine tune on top.
I have been running evals versus other models on my own data. For my use case, Jev comes out on top at least 2/3 of the time. This is purely about accuracy. For cost and speed, it is on top 100% of the time. I agree that people should test their own data against it before blindly assuming that it will get better outcomes.