As I said, having choices is, of course, the best. Kagi can certainly achieve both high-speed and high-quality (I mean without lowering the existing quality at all—not even a little bit), but it’s hard to avoid an increase in costs.
Of course, this is just my personal choice based on my experience. If you prioritize speed, as Giorgio mentioned, you can choice translation with speed/quality tradeoff.
My understanding of non-human translation is this: when the user doesn’t have a basic understanding of the language being translated, quality becomes extremely important because it’s very difficult to tell whether the translation is accurate or not—unless you’re very familiar with the field of the content being translated.