BLAST RADIUS

Open-weight & owned stack · Quantization efficiency

Qwen3.8 Next Runs On Six V100 GPUs

Sep 20, 2026

A user reported running Qwen3.8 Next on six V100 GPUs at 22.59 tokens per second without speculative decoding and 42.70 with it enabled.

Read the original at reddit.comOpens the publisher's site in a new tab

More in Open-weight & owned stack