Comment by wolfgangK

Isn't ChatGPT benchmaxxing, then ? Responding "hmm…" isn't actually responding and latency should time to first relevant phoneme.