#Rate limit bypass
1 messages · Page 1 of 1 (latest)
Hi @frozen robin, what else do we know about this? Does it happen with all the models? Can you share more information?
I'll message you when I get back in, but yeah, it's completely model agnostic and is to do with exactly how requests are handled by the server and how tokens are handled. I mean, I've got better words when I get back home.
Here's a clue. There is absolutely no client side rate throttling. And you can submit queries via the interface or directly to the query serving server thing.
tokens can be rotated around and I mean I don't even see my point going down when I test it to be honest
But yeah, I'll send you the full proof of concept. Well, actually less proof of concept, more weaponized version of this when I get back.