RCreddit.com
18
·开发者社区 · RSS
Whats the catch with SwiReasoning?
I just heard about SwiReasoning and tried it out on Qwen 3.6 27b and im kinda surprised.
Its answers are more on point and it solves questions aloooot quicker.
It seems a bit slower in t/s but the amount of tokens it needs is so much lower, it feels faster.
Anybody else tried it? Wheres the catch? Its a 9 month old technique, why isnt it all over the place?
Sources:
- https://github.com/sdc17/SwiReasoning
- https://github.com/Antonbe1b/swireasoning-llamacpp
原始关键词#swireasoning#catch#whats