Skip to content

Qwen3.8 Weights Released

modelscope.cn
31 pointsapitman4 comments
On HN

Comments

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.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.