Hacker News new | past | comments | ask | show | jobs | submit
anana_ | on: Qwen 3.8 27B
As was the case with GLM 5.3, it seems that there is still much juice to be squeezed from post-training