BLAST RADIUS

Open-weight & owned stack · Quantization efficiency

Qwen3.8 Flash Next Shows High Local Throughput

Sep 29, 2026

A user reported serving quantized Qwen3.8-Flash-Next at over 150 tokens per second singly and roughly 100 tokens per second across concurrent streams.

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

More in Open-weight & owned stack

Qwen releases full-duplex realtime audio modelSep 29Companies Shift Model Usage Toward Open-Weight AISep 28NVIDIA Releases Open-Weight Competitive-Coding ModelSep 28NVIDIA CEO calls AI distillation competitionSep 29AutoTrust releases open JEV-27B decision modelSep 29