Magnum manages to impress again
#1
by
Ainonake
- opened
This feels smarter than other models in v4 series. I don't know what you did, but it seems like you managed to fix boring llama 3.3 slop, without hurting intilligence.
It defintely is smarter than behemoth and other stuff on largestral - it reacts to prompts and instructions much better (behemoth would often just ignore them).
Seems like people have found some secret sauce to fixing llama 3 models? Because recent sao's tunes were a decent stepup from earlier Eureales. This model feels much smarter than Cirrus though.
This is my favourite model now