deepseek-v4.1VIDEO
4 张 2019 年老卡跑 V4.1 Flash：峰值 314 tok/s
用户用 4 张 2019 年老 GPU 跑原生未改权重的 V4.1 Flash，真实编程任务峰值解码 314 tok/s，agent 平均 200 tok/s。
Qubitium
@qubitium7 天前收录原推发布于 2026.10.24
Qubitium @qubitium原推发布于 2026.10.24
DeepSeek v4.1 Flash 1017 just reached batch 1 peak decode of 314 t/s on real coding work, not simulation. Avg 200 t/s on my coding agents. Setup is 4x 2019 era gpu. Full native, unmodified weights. Imagine what I can do with 2025 gpus. 😎
x.com/qubitium/status/2113817933424312680采集于 2026.10.24 12:05