artificialbuilds

Alibaba Cloud· Released Aug 26

Education with Qwen3.8-Flash-Next

Strata makes Flash-Next fast in 12GB VRAM (Japanese)Sep 26

ハカセアイ

@ai_hakase_ · Sep 26

Japanese explainer of the lazy dequantisation and expert cache design, with about 65 tok/s.

More from Qwen3.8-Flash-Next

Strata: one-click engine for running the 125B Flash-Next on a gaming PCon GitHub:Niko1221· 16.8K stars16.8K stars
Flash-Next on a single DGX Spark (TP=1)on GitHub:MiaAI-Lab· 594 stars594 stars
Halogen: fastest Flash-Next server for AMD Strix Haloon GitHub:peonist-ai· 876 stars876 stars
GVS5H: five Qwen3.8 models matching a frontier model on LiveCodeBench Hardon GitHub:slee-persis· 736 stars736 stars