Qwen3.8 Weights Released

(modelscope.cn)

26 points | by apitman 1 hour ago

3 comments

throwfaraway4 1 hour ago

https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B
13hours 1 hour 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 1 hour 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 1 hour ago

Next up: https://modelscope.cn/models/Qwen/Qwen3.8-27B
Probably easier to run for everyone.