Local AI community urges Qwen to bring back a 35B-class MoE for low-VRAM GPUs

julianharris · x · 2026-09-27

A developer is petitioning the Qwen team on behalf of the consumer-GPU local AI community for a new 35B MoE model, arguing Qwen-3.6-35B-A3B is the only model in that range that gave 4GB-16GB VRAM users a real seat in local AI. He also hopes for a Qwen4 n-gram feature and a 9B variant.

Original post →

More from Models

Models channel →