Qwen3-Next: Towards Ultimate Training and Inference Efficiency (qwen.ai)

16 points by bratao 10mo ago ↗ HN

1 comment

[ 3.5 ms ] story [ 11.2 ms ] thread

derac 10mo ago ↗

This seems very promising, they are matching Qwen3-235B-A22B-Instruct-2507. It should run on 8-12GB vram with 64 gb ram decently well.