deepseek-v4.1IMAGE

Maya 优化：V4.1 Flash 1017 预填充提速近 3 倍

同一套 3×V100 32GB 上跑同一份 V4.1 Flash 1017 量化，Maya 更新后预填充从约 420 tok/s 升到约 1400 tok/s，解码速度未降。

@ItsmeAjayKV7 天前收录原推发布于 2026.10.24

AJ @ItsmeAjayKV原推发布于 2026.10.24

Heads up! Maya just got faster ! On same 3x Tesla V100 32GB + DDR4 system running same DeepSeek-v4.1-Flash-1017 quant just got a whole lot faster. Prefill is improved by almost 3x, before: ~420 tok/s after: ~1400 tok/s That's huge, all while decode is not going down !! @PeasantSmith is moving fast with Maya, much appreciated.

这条是引用着另一条写的：x.com/peasantsmith/status/2113730024125927571

x.com/itsmeajaykv/status/2113773463148081488采集于 2026.10.24 12:06