Opus 4.5 can oneshot 600 line Mastodon bot

It's truly joever

Nope, it still made a bug, but wow, what a progress from gpt-3.5

wtf, i don't think that's a bug, i think that's the model shitting the bed

yep, gemma-12b heretic is too retarded

and qwen3-vl heretic just stops thinking, wtf
{'role': 'assistant', 'reasoning_content': 'Okay, the user wants me to role', 'content': ''}

32b 6k seem to work fine

Way too slow if you can't fit it in VRAM.
Letting the GPU do all the processing and letting it overflow into RAM is faster than having llama.cpp split it between CPU and GPU

qwen3-vl-8b abliterated seems to work well

interesting though, seems like prompt caching can poison rhe context and start things like typing in all caps