deepseek-v4.1IMAGE
DeepSeek 开源生产推理引擎 DeepInfer，默认跑 V4.1-Flash-1017
DeepInfer 支持 NVIDIA SM100 与华为昇腾 950 NPU，默认模型为 1M 上下文的 V4.1-Flash-1017，带确定性采样与结构化输出。
Decoded
@Decoded085 天前收录原推发布于 2026.10.25
Decoded @Decoded08原推发布于 2026.10.25
🚀 DeepSeek just open-sourced DeepInfer, the inference engine behind its own production services. Runs on NVIDIA SM100 GPUs & Huawei Ascend 950 NPUs, defaults to DeepSeek-V4.1-Flash-1017 (1M context), with deterministic sampling & structured outputs. #AI #OpenSource
x.com/decoded08/status/2114202336813875506采集于 2026.10.25 12:00