bgg1996

/

Melinoe-Next-80B-A3B-Thinking-v0

like 3

hi!
i love the model.
would you consider doing a qwen3-next-instruct based version?
im running these on cpu and the reasoning takes up a lot of time :p also, the reasoning sometimes doesnt look much different from the final output, or the model seems to just forget to actually think and just responds, but then the text ends up in the reasoning block and that messes with its visibility for future turns and stuff.
so it seems like it almost wants to be an instruct model in some sense haha.
i understand this stuff is not easy or free though, so no pressure of course, and thank you for the model!