読書リサーチレーダー投資フレームワーク
ログイン / 新規登録
← 読解に戻る
トピックDisconfirmAI 編集部更新日 2026-11-07 19:16

DeepSeek推理框架

4 ステップ5 件4 ソース

中国AI公司DeepSeek于2026年10月20日正式开源其内部生产级推理框架DeepInfer及配套组件deepseek-recipe,旨在降低大规模模型部署的算力与能源成本。该框架支持英伟达Blackwell(SM100)和华为昇腾950芯片,目前仅适配DeepSeek-V4.1-Flash模型。DeepSeek披露单集群日吞吐155万亿Token,非高峰时段每GPU小时营收38美元,引发社区对高利润率与回本周期的讨论。然而,核心KV缓存管理与负载均衡代码因耦合内部基础设施暂未公开,且vLLM团队指出其在GB200上的解码吞吐量差距在6%以内,质疑其效率壁垒是否不可逾越。

経緯

  1. 2026-10-20

    DeepSeek宣布开源内部推理框架DeepInfer及编解码组件deepseek-recipe,支持英伟达与华为平台。

    ithome.com ↗
  2. 2026-10-20

    SiliconANGLE报道DeepSeek开源DeepInfer,指出部分核心调度代码暂未公开,并提及美国国土安全部此前要求联邦机构移除DeepSeek产品。

    siliconangle.com ↗
  3. 2026-10-21

    Latent Space发布AINews深度解析,披露DeepSeek声称的单集群日吞吐155万亿Token及每GPU小时38美元营收数据,引发关于利润率与回本周期的争议。

    latent.space ↗
  4. 2026-10-22

    t3n.de报道DeepInfer通过优化推理阶段计算分配来降低企业自托管模型的算力和能源成本,但仅适配特定高端芯片。

    t3n.de ↗

未解決の問題

  • DeepInfer未公开的KV缓存管理和负载均衡代码何时能完全解耦并开源?
  • vLLM等现有开源引擎能否通过配置优化达到DeepInfer宣称的效率水平?
  • DeepSeek披露的经济数据(如38美元/GPU小时营收)在不同工作负载和硬件成本下的实际利润率是多少?

ソース

ログイン / 新規登録
読書リサーチレーダー投資フレームワーク
  • ithome.comIT之家:DeepSeek开源内部推理框架DeepInfer↗
  • siliconangle.comSiliconANGLE: DeepSeek open-sources DeepInfer↗
  • latent.spaceLatent Space: DeepSeek DeepInfer Economics Analysis↗
  • github.comGitHub: deepseek-ai/DeepInfer Repository↗

このトピックでのカバー範囲

  • Latent Space AINews Analyzes DeepSeek's Open-Sourced Inference Engine DeepInfer: Details on 155T Daily Tokens and $38/GPU-Hour Revenue2026-10-23ニュース

    This AINews episode deeply reports on Chinese AI firm DeepSeek's open-sourcing of its production-grade inference framework, DeepInfer. The program goes beyond technical architecture (such as disaggregated prefill/decode and SWA bounded replay) to analyze disclosed economic data: $38 revenue per GPU-hour at off-peak pricing, alongside community debates on derived margins and payback periods. It is a valuable resource for readers focused on AI infrastructure cost structures and commercial viability.

  • DeepSeek releases open-source inference framework Deepinfer to cut self-hosting costs2026-10-22ニュース

    DeepSeek launched Deepinfer, an open-source tool designed to reduce compute and energy costs for enterprises self-hosting AI models, currently limited to specific high-end hardware.

  • DeepSeek Open-Sources DeepInfer: Compatible with Nvidia and Huawei Chips, Claims $38/Hour Revenue Per GPU2026-10-21ニュース

    DeepSeek releases DeepInfer to optimize V4.1-Flash model efficiency on Nvidia and Huawei hardware, challenging closed-source rivals through open ecosystem.

  • DeepSeek Open-Sources DeepInfer, Its Production Inference Framework Supporting Nvidia Blackwell and Huawei Ascend2026-10-21ニュース

    DeepSeek has open-sourced DeepInfer, its production-grade inference framework featuring speculative decoding and prefill optimizations, with support for Nvidia Blackwell and Huawei Ascend 950 chips, though some core scheduling code remains closed.

  • DeepSeek Open-Sources Internal Inference Framework DeepInfer for High-Throughput Deployment2026-10-20ニュース

    DeepSeek has fully open-sourced its internal inference framework DeepInfer and the accompanying tokenization component deepseek-recipe, supporting both NVIDIA and Huawei Ascend platforms to help developers reduce compute costs for large-scale model deployment.