Interesting that you probably meant performance as in latency (and probably we'll see over time, availability too)

But the first thought that jumped to my mind was you meant "performance" as in "quality" -> the believability/humanity of the "performance" of the chatbot.

It was a silly misunderstanding on my part but I suspect it's indicative of an important point: This product war won't be won on latency - people are happy to wait seconds for most of the scenarios, and possibly even MINUTES for many of them.