[rightbyte]

Running models locally brings my beefy rig to the knees for about half a minute for each querry for smaller models. Answering querries has to be expensive too?