Hacker News
new
top
best
ask
show
job
Qwen3.8 Weights Released
(
modelscope.cn
)
26 points
by
apitman
3 hours ago
3 comments
throwfaraway4
3 hours ago
https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B
13hours
3 hours ago
Anyone know if a MoE model that can run on a gpu with 16GB VRAM (like 3.6-35B-A3B) is in the pipeline as well, or only the dense 27B one?
petu
3 hours ago
Whoever runs their Twitter wouldn't confirm / promise anything beyond 27B, but didn't say no either. So maybe in a month, maybe not.
spwa4
3 hours ago
Next up:
https://modelscope.cn/models/Qwen/Qwen3.8-27B
Probably easier to run for everyone.