[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-prompt-caching-90pct-token-cost":3,"topics-all":36,"news-related-2fa66657-afbb-4f03-849a-f420f42cf2ab":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"2fa66657-afbb-4f03-849a-f420f42cf2ab","Prompt Caching：LLM推理成本削减90%的隐藏利器","**LLM推理成本的新解法：Prompt Caching如何实现90%token费用减免**\n\n调用大模型API时，每次请求都要重复发送相同的系统提示词和上下文——这是一种隐性浪费。Prompt Caching（上下文缓存）正在改变这一现状。\n\n技术原理并不复杂。LLM推理分为两个阶段：Pre-fill（处理完整提示词生成首个token，计算密集型）和Decoding（自回归逐token生成，内存带宽密集型）。当多个请求共享相同的系统提示词或基础上下文时，将这部分内容缓存下来复用，就能省去重复计算的开销。\n\n根据OpenAI官方文档，Prompt Caching可实现最高80%的延迟降低和90%的输入token成本削减。Anthropic、Google Gemini等主流厂商都已支持这一特性，关键在于如何设计提示词结构以最大化缓存命中率。\n\n但缓存策略并非万能。共享的系统前缀必须出现在提示词的固定位置，缓存失效后的冷启动延迟反而更高。开发者需要在缓存命中率与提示词灵活性之间找到平衡点。\n\n对于需要频繁调用大模型的AI应用——无论是RAG系统、代码生成工具还是多轮对话Agent——Prompt Caching都是值得关注的基础设施级优化。在模型能力差距逐渐收窄的当下，推理效率正在成为新的竞争维度。","https:\u002F\u002Ftowardsdatascience.com\u002Fwhy-care-about-promp-caching-in-llms\u002F","49852722-6459-439c-80a7-dc2726bd3aa0",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"7ac06d8e-b074-4147-abfc-ffaa4c6b8744","ai-efficiency",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"ad7220f1-c43a-4949-a338-cd2bb18586ff","en","Prompt Caching: The Hidden Lever for Cutting LLM Inference Cost by 90%","When calling a large-model API, the same system prompt and context get re-sent on every request — a hidden tax. Prompt Caching is changing that.\n\nThe technical principle is straightforward. LLM inference has two phases: pre-fill (processing the full prompt to generate the first token, compute-heavy) and decoding (autoregressive token-by-token generation, memory-bandwidth-heavy). When multiple requests share an identical system prompt or base context, caching that portion for reuse eliminates redundant computation.\n\nAccording to OpenAI's official documentation, Prompt Caching can deliver up to 80% latency reduction and 90% input-token cost cut. Anthropic, Google Gemini, and other major vendors have all added support; the key is how to design your prompt structure to maximize cache-hit rates.\n\nThat said, caching strategy is not a silver bullet. The shared system prefix must occupy a fixed position in the prompt, and a cold-start after a cache miss actually carries higher latency. Developers need to strike a balance between cache hit rate and prompt flexibility.\n\nFor any AI application that calls LLMs frequently — whether RAG systems, code generation tools, or multi-turn dialogue agents — Prompt Caching is a piece of infrastructure-level optimization worth attention. As model-capability gaps narrow, inference efficiency is becoming the new competitive axis.","prompt-caching-90pct-token-cost","2026-05-26T01:10:00Z","2026-05-26T01:09:27.827379Z","2026-08-19T02:08:40.142862Z",true,"agent",199,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"748a4486-34e7-4215-b515-7eb56b3258c5","TwELL：Sakana AI与NVIDIA联合提出稀疏LLM推理加速20%，解决GPU批处理落地难题","sakana-nvidia-twell-20pct-sparse-batch-gemm","2026-05-30T08:20:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"73f4d31e-745a-4bba-8a0f-38e7564966de","Sakana AI 提出 99% 稀疏性Transformer：在前馈层动刀革新LLM效率","sakana-99pct-sparse-ffn-transformer","2026-05-16T19:04:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"137ce22e-389d-47fb-8219-42ca53d6e916","Qwen 3.6 27B 重磅更新：MTP 技术让本地推理提速 2.5 倍","qwen-3-6-27b-mtp-local-2-5x","2026-05-16T01:01:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"4556940e-6456-43ce-b9b4-a0a7fa7a5865","MIT 新方法：自适应草稿模型将推理 LLM 训练速度提升 2-3 倍","mit-adaptive-draft-speculative-train-2-3x","2026-05-15T02:05:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"e4fd45e9-e0fd-4839-973e-909a442ce5ff","DeepSeek V3.2稀疏注意力：如何将长上下文推理成本砍半","deepseek-v3-2-dsa-sparse-attention-50pct-cost-cut","2026-05-01T10:15:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"810251ee-b8bf-4fef-a9d8-e167c22ae4c5","BoostLoRA：梯度增强让低秩适配器「自我进化」，小参数也能有大表达","boostlora-gradient-boosting-lora-residual","2026-05-01T05:10:00+00:00"]