DeepSeek开源推理引擎DeepInfer
DeepSeek just open-sourced DeepInfer, the inference engine behind its production…
DeepSeek开源生产级推理引擎，涵盖DSpark与Bounded Replay等核心加速细节，对做高性能部署的工程团队极具参考价值。
DeepSeek just open-sourced DeepInfer, the inference engine behind its production services, under the Apache 2.0 license.
DeepSeek 刚刚以 Apache 2.0 许可证开源了 DeepInfer，这是其生产服务背后的推理引擎。
and it got 6.3K+ stars in 8 hours
它在 8 小时内获得了 6.3K+ 颗星
The release runs only DeepSeek-V4.1-Flash, on NVIDIA SM100 GPUs or Huawei Ascend 950 chips, and covers offline inference rather than a complete serving stack.
该版本仅支持 DeepSeek-V4.1-Flash，运行在 NVIDIA SM100 GPU 或华为昇腾 950 芯片上，涵盖离线推理而非完整的推理服务栈。
It omits the parts that turn an engine into a service: a KV cache system, prefill and decode disaggregation, and load balancing across replicas.
它省略了将引擎转化为服务的部分：KV 缓存系统、预填充与解码分离，以及跨副本的负载均衡。
What ships is the speed layer, including DSpark, which drafts up to 5 tokens ahead and uses confidence estimates to schedule their verification.
发布的内容是速度层，包括 DSpark，它能提前推导多达 5 个 token，并利用置信度估计来调度它们的验证。
Bounded replay trims prefill by recomputing only a 140-token suffix sized to cover the sliding attention window.
有界重放（Bounded replay）通过仅重新计算覆盖滑动注意力窗口的 140 个 token 后缀来缩减预填充开销。
A companion library, deepseek-recipe, will turn Chat Completions, Responses and Messages requests into DeepSeek V4 and V4.1 prompts that match the official API’s behaviour.
一个配套库 deepseek-recipe 会将 Chat Completions、Responses 和 Messages 请求转换为符合官方 API 行为的 DeepSeek V4 和 V4.1 提示词。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力