GPT 5.2 Extended thinking webchat has unworkably slow token (4 tps) generation

If you can time token generation after start of first token (so not including thinking time, which is a different kettle of fish), that would help give some specifics on the problem. The modes are Auto, Instant, Thinking, and Extended Thinking on my web client. When I saw the problem, it only occurred on Thinking and Extended Thinking.