Qwen3.8 Weights Released
modelscope.cn
3 threads
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?
Whoever runs their Twitter wouldn't confirm / promise anything beyond 27B, but didn't say no either. So maybe in a month, maybe not.
Next up: https://modelscope.cn/models/Qwen/Qwen3.8-27B
Probably easier to run for everyone.