Comment by mendeza

what is the throughput for gpt-oss, 1 token every 2 seconds is really slow, but understandable because you are moving cache to disk