@Blizzr_@cuntradiation I’ve seen this before way too many times. This is a small LLM (old 1b models or under) going crackshit insane and being unable to respond to anything else. If anything, you should use something like liquid or locally with a gemma or LLaMa model