NexBench 实测：DeepSeek V4.1 Flash 1017 在 18 个模型中得分最高

MindFort AI @mindfort原推发布于 2026.10.23

在 X 上打开原推

We evaluated DeepSeek-V4.1-Flash-1017 on NexBench, producing the highest score out of 18 total models. NexBench provides signals of a model's ability to discover, reason about, and execute real exploits against a real target in a messy environment, while measuring price-to-performance ratio, token efficiency, and more. See our thread for more findings (1/4)