<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>有出处的 AI 资讯</title><link>https://aisubhub.top/zh/news/</link><description>追踪产品发布、模型更新与研究，保留一手来源，讲清实际影响和适用边界。</description><language>zh</language><lastBuildDate>Tue, 20 Oct 2026 10:29:03 GMT</lastBuildDate><atom:link href="https://aisubhub.top/zh/news/feed.xml" rel="self" type="application/rss+xml"/><item><title>微软主权AI更新：哪些已经可用，哪些仍是预览或后续计划</title><link>https://aisubhub.top/zh/news/microsoft-sovereign-ai-local-controls/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/microsoft-sovereign-ai-local-controls/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>10月20日公告把本地AI、开发工具与运维访问控制放在一起，但已经可用、区域推广、预览和Ignite后续发布必须分清。</description><source url="https://blogs.microsoft.com/blog/2026/10/20/advancing-digital-sovereignty-with-greater-control-and-choice/">原始来源</source></item><item><title>BSC为西班牙AI初创企业打开算力通道：免费参与不等于无限资源</title><link>https://aisubhub.top/zh/news/bsc-ai-factory-inception-startup-access/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/bsc-ai-factory-inception-startup-access/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>加速器已正式运行，结合技术支持与商业辅导；入选条件、算力分配和示范项目阶段仍需区分。</description><source url="https://blogs.nvidia.com/blog/spain-startups-barcelona-supercomputing-center/">原始来源</source></item><item><title>AWS为Bedrock中的Claude增加结构化输出</title><link>https://aisubhub.top/zh/news/bedrock-claude-structured-outputs/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/bedrock-claude-structured-outputs/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>JSON结构与工具参数可受约束，但事实核验及不受支持的规则仍由应用负责。</description><source url="https://aws.amazon.com/blogs/machine-learning/structured-outputs-for-anthropic-claude-models-on-amazon-bedrock/">原始来源</source></item><item><title>SageMaker Studio 简化新账号入门，额度与 GPU 条件仍需分清</title><link>https://aisubhub.top/zh/news/sagemaker-studio-new-account-onboarding/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/sagemaker-studio-new-account-onboarding/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>AWS 把 Studio 接入更短的注册流程，降低开始实验的门槛，但没有提供无限云计算。</description><source url="https://aws.amazon.com/blogs/machine-learning/developers-can-now-access-sagemaker-studio-through-the-new-aws-experience/">原始来源</source></item><item><title>德国与NVIDIA提出工业AI合作方向</title><link>https://aisubhub.top/zh/news/nvidia-germany-industrial-ai-cooperation/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/nvidia-germany-industrial-ai-cooperation/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>10月19日公司报道介绍意向声明与计划投资，不代表部署或交割完成。</description><source url="https://blogs.nvidia.com/blog/germany-wants-to-lead-the-world-in-industrial-ai-how-nvidia-will-help/">原始来源</source></item><item><title>AWS详解Bedrock中Pegasus 1.5的视频分析</title><link>https://aisubhub.top/zh/news/pegasus-15-bedrock-video-metadata/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/pegasus-15-bedrock-video-metadata/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>10月19日说明稿回顾10月9日上线，介绍时间戳JSON与两小时视频处理。</description><source url="https://aws.amazon.com/blogs/machine-learning/pegasus-1-5-from-twelvelabs-is-now-available-in-amazon-bedrock/">原始来源</source></item><item><title>Deepgram的Trainium案例：模型、内核与服务成本需分开看</title><link>https://aisubhub.top/zh/news/deepgram-nova3-trainium2-batch-inference/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/deepgram-nova3-trainium2-batch-inference/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>AWS报告Nova-3离线转写经定制优化后每美元处理更多音频，但不能外推为所有语音服务的降价承诺。</description><source url="https://aws.amazon.com/blogs/machine-learning/how-deepgram-optimized-voice-ai-at-scale-with-aws-trainium/">原始来源</source></item><item><title>vLLM两种KV传输策略：不同延迟瓶颈，不是叠加加速</title><link>https://aisubhub.top/zh/news/vllm-kv-transfer-latency-and-multi-turn-reuse/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/vllm-kv-transfer-latency-and-multi-turn-reuse/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>AWS两篇指南对照push交接与双向缓存复用，测试条件不同，不能把倍数相乘。</description><source url="https://aws.amazon.com/blogs/machine-learning/push-based-kv-transfer-to-improve-ttft-and-tpot-for-disaggregated-inference-in-vllm/">原始来源</source></item><item><title>NVIDIA把主权广播框架与媒体检索提速放在同一议程</title><link>https://aisubhub.top/zh/news/ebu-sovereign-media-frameio-agentic-search/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/ebu-sovereign-media-frameio-agentic-search/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>EBU合作处理治理与基础设施，Adobe Frame.io的性能数据则属于另一条技术路径。</description><source url="https://blogs.nvidia.com/blog/media-agentic-security-european-public-service-broadcasting-global-content-production/">原始来源</source></item><item><title>微软主张持续监管企业内部AI agent</title><link>https://aisubhub.top/zh/news/microsoft-agentic-insider-2026-10-19/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/microsoft-agentic-insider-2026-10-19/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>微软10月19日文章将安全重点延伸至企业自行部署、拥有工具和数据权限的内部agent。</description><source url="https://blogs.microsoft.com/blog/2026/10/19/securing-the-agentic-insider/">原始来源</source></item><item><title>OpenAI 公布低资源编程语言评测进展</title><link>https://aisubhub.top/zh/news/openai-low-resource-coding-evaluations/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/openai-low-resource-coding-evaluations/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>新评测关注 SQLite 实现与保持行为一致的 COBOL 迁移，结果来自 OpenAI 官方报告。</description><source url="https://openai.com/index/low-resource-programming-languages/">原始来源</source></item><item><title>Qwen Code v0.25.3增加隔离审查与执行控制</title><link>https://aisubhub.top/zh/news/qwen-code-v0253-2026-10-19/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/qwen-code-v0253-2026-10-19/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>10月19日稳定版更新聚焦托管会话、工具配置和工作区边界，并非新模型或会员发布。</description><source url="https://github.com/QwenLM/qwen-code/releases/tag/v0.25.3">原始来源</source></item><item><title>Hugging Face介绍TorchNeuron原生Trainium支持</title><link>https://aisubhub.top/zh/news/hugging-face-torchneuron-2026-10-19/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/hugging-face-torchneuron-2026-10-19/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>10月19日文章说明常用模型库如何直接使用新后端，并预告10月20日开启公开测试。</description><source url="https://huggingface.co/blog/hf-torchneuron">原始来源</source></item><item><title>RTX Spark上市把Windows笔记本与分阶段本地AI生态连接起来</title><link>https://aisubhub.top/zh/news/rtx-spark-windows-launch-software-stages/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/rtx-spark-windows-launch-software-stages/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>三篇NVIDIA文章属于同一次产品发布，硬件已售不等于全部软件功能均已正式可用或免费。</description><source url="https://blogs.nvidia.com/blog/local-ai-rtx-spark-windows-laptops-celebration/">原始来源</source></item><item><title>DeepSeek通过既有API别名更新Flash模型</title><link>https://aisubhub.top/zh/news/deepseek-flash-1017-update/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/deepseek-flash-1017-update/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>10月17日记录的是模型升级，API调用与定价未变，不是新增个人会员。</description><source url="https://api-docs.deepseek.com/updates/">原始来源</source></item><item><title>AWS 用九阶段闭环检查 Agent 质量，不只看服务在线</title><link>https://aisubhub.top/zh/news/agentcore-nine-stage-quality-evaluation/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/agentcore-nine-stage-quality-evaluation/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>内部销售助手案例把业务评估、发布回放与线上监测相连，阈值仍取决于具体负载。</description><source url="https://aws.amazon.com/blogs/machine-learning/building-trust-in-ai-agents-a-production-ready-evaluation-framework-with-amazon-bedrock-agentcore/">原始来源</source></item><item><title>Anthropic 研究本公司工作变化与初级技术招聘</title><link>https://aisubhub.top/zh/news/anthropic-internal-employment-study/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/anthropic-internal-employment-study/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>内部报告讨论代理采用与工作方式的关联，并提醒因果解释及全社会外推存在限制。</description><source url="https://www.anthropic.com/research/employment-at-anthropic">原始来源</source></item><item><title>AWS解读Reflection Beam部署：活跃参数不等于所需显存</title><link>https://aisubhub.top/zh/news/reflection-beam-aws-deployment-guide/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/reflection-beam-aws-deployment-guide/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>10月16日指南说明5010亿参数编程模型的托管与HyperPod路径，没有统一价格或效果保证。</description><source url="https://aws.amazon.com/blogs/machine-learning/introducing-reflection-beam-an-open-weight-model-for-agentic-coding-on-aws/">原始来源</source></item><item><title>a16z 作者分析：AI 购物为何仍面临阻力</title><link>https://aisubhub.top/zh/news/a16z-agentic-commerce-permissions/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/a16z-agentic-commerce-permissions/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>Moses Sternstein 区分辅助选购与委托结账，并提醒部分引用数据存在时效限制。</description><source url="https://www.a16z.news/p/this-week-in-charts-agents-denied">原始来源</source></item><item><title>Cursor 云端 Agent 新增可分享画布与测试版环境 API</title><link>https://aisubhub.top/zh/news/cursor-cloud-agent-canvases-environments-2026-10-16/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/cursor-cloud-agent-canvases-environments-2026-10-16/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>10月16日短更新将交互成果分享、程序化环境管理及更大的仓库数量上限放到同一工作流。</description><source url="https://cursor.com/changelog/cloud-agent-canvases-and-environments">原始来源</source></item><item><title>Elorian 发布 Glimpse 视觉推理研究预览</title><link>https://aisubhub.top/zh/news/elorian-glimpse-research-preview/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/elorian-glimpse-research-preview/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>原文介绍 API 与聊天入口，并明确区分模型录制输出与用于说明的参考答案。</description><source url="https://elorian.ai/blog/introducing-glimpse">原始来源</source></item><item><title>Rosalind 学术计划面向美国高校实验室，访问须审批</title><link>https://aisubhub.top/zh/news/rosalind-academic-plan-eligibility/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/rosalind-academic-plan-eligibility/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>符合条件的实验室可申请赞助席位；GPT-Rosalind 需要审核，计划不含 API 访问。</description><source url="https://openai.com/index/introducing-the-rosalind-academic-plan/">原始来源</source></item><item><title>微软研究院发布Magnetar可验证计算研究</title><link>https://aisubhub.top/zh/news/microsoft-research-magnetar-2026-10-15/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/microsoft-research-magnetar-2026-10-15/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>10月15日研究文章介绍基于RISC-V的零知识虚拟机，并给出特定基准中的性能结果。</description><source url="https://www.microsoft.com/en-us/research/blog/proof-based-general-purpose-computing-with-magnetar/">原始来源</source></item><item><title>AWS 的 Trust in Depth：让 Agent 决策，让外部控制守边界</title><link>https://aisubhub.top/zh/news/aws-trust-in-depth-agent-boundaries/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/aws-trust-in-depth-agent-boundaries/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>四层方法连接任务范围、身份、执行约束和证据；形式化证明仅适用于明确规则与假设。</description><source url="https://aws.amazon.com/blogs/machine-learning/trust-in-depth-giving-agents-more-autonomy-without-losing-control/">原始来源</source></item><item><title>Nemotron 3.5分清多模态智能体与小内存本地部署</title><link>https://aisubhub.top/zh/news/nemotron-35-super-vl-bolt-release/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/nemotron-35-super-vl-bolt-release/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>120B参数Super VL与24GB版Bolt同期发布，性能、内存和上线时间必须按具体型号判断。</description><source url="https://blogs.nvidia.com/blog/nemotron-3-5-super-vl/">原始来源</source></item><item><title>NVIDIA欧洲科研承诺：开放教材与有条件的算力通道</title><link>https://aisubhub.top/zh/news/nvidia-europe-science-access-pledge/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/nvidia-europe-science-access-pledge/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>10月15日介绍的支持包括学习资源与科研路径，不等于自动分配GPU或统一商业许可。</description><source url="https://blogs.nvidia.com/blog/ec-scientific-computing-pledge/">原始来源</source></item><item><title>NVIDIA北欧更新：已发布系统与未来AI容量要分清</title><link>https://aisubhub.top/zh/news/nvidia-nordic-infrastructure-project-stages/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/nvidia-nordic-infrastructure-project-stages/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>区域公告连接共享超算、应用优化和扶持计划，却不代表它们构成人人可用的统一服务。</description><source url="https://blogs.nvidia.com/blog/nordics-technology-ecosystem/">原始来源</source></item><item><title>MAI-Base-2训练详解：效率提升与模型能否使用，要分别核实</title><link>https://aisubhub.top/zh/news/mai-base-2-pretraining-methods/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/mai-base-2-pretraining-methods/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>微软公开2.8万亿参数基座的训练方法，涉及重复数据缩放、专家调度与优化器；效率数字不等于现成的聊天产品或订阅权益。</description><source url="https://microsoft.ai/news/frontier-pre-training-at-microsoft-ai/">原始来源</source></item><item><title>AgentCore 文档处理复用已验证指令，不是重新训练模型权重</title><link>https://aisubhub.top/zh/news/agentcore-document-processing-instruction-reuse/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/agentcore-document-processing-instruction-reuse/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>AWS 样例用八个 Agent 配合业务校验，第二次更快与月成本都只对应特定条件。</description><source url="https://aws.amazon.com/blogs/machine-learning/build-self-improving-document-processing-using-amazon-bedrock-agentcore/">原始来源</source></item><item><title>AgentCore 演示文稿样例是 RAG 工作流，不是四美分幻灯片服务</title><link>https://aisubhub.top/zh/news/agentcore-rag-presentation-generation/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/agentcore-rag-presentation-generation/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>AWS 内测六页结果有模型成本、异步交付和 OpenSearch 网络条件，不能当全包价。</description><source url="https://aws.amazon.com/blogs/machine-learning/generate-presentations-with-amazon-bedrock-agentcore/">原始来源</source></item><item><title>Tamnoon 展示 Tami 如何把云调查与获批修复分开</title><link>https://aisubhub.top/zh/news/tamnoon-bedrock-governed-cloud-remediation/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/tamnoon-bedrock-governed-cloud-remediation/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>10 月 14 日 AWS 案例披露四类智能体、保守安全评级与受控执行；改善幅度属于厂商报告，不能当作保证。</description><source url="https://aws.amazon.com/blogs/machine-learning/how-tamnoon-built-ai-powered-cloud-security-remediation-with-amazon-bedrock/">原始来源</source></item><item><title>AWS 给出 JumpStart 模型部署、评估与清理的可重复流程</title><link>https://aisubhub.top/zh/news/sagemaker-jumpstart-repeatable-model-evaluation/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/sagemaker-jumpstart-repeatable-model-evaluation/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>10 月 14 日指南串起模型发现、样例请求和端点删除；示例速度不能当成通用性能榜单。</description><source url="https://aws.amazon.com/blogs/machine-learning/deploy-and-evaluate-models-with-amazon-sagemaker-jumpstart/">原始来源</source></item><item><title>NVIDIA拟投资五家英国初创企业，覆盖生物研究与量子系统</title><link>https://aisubhub.top/zh/news/nvidia-uk-startup-planned-investments-2026/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/nvidia-uk-startup-planned-investments-2026/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>10月14日公告列出新增投资计划及既有应用，但未披露各笔金额或融资完成情况。</description><source url="https://blogs.nvidia.com/blog/uk-ecosystem-deployment/">原始来源</source></item><item><title>欧洲城市AI项目状态不一：已部署、试点和成本目标需分开</title><link>https://aisubhub.top/zh/news/nvidia-european-city-agents-project-stages/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/nvidia-european-city-agents-project-stages/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>NVIDIA介绍本地智能体与数字孪生，但不同市政及港口案例的阶段和证据差异明显。</description><source url="https://blogs.nvidia.com/blog/european-cities-agents-urban-operations/">原始来源</source></item><item><title>Bedrock 私有报价要先激活授权，协商价格才会适用</title><link>https://aisubhub.top/zh/news/bedrock-private-offer-managed-entitlements/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/bedrock-private-offer-managed-entitlements/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>AWS 对比一次性审阅脚本与卖方白名单自动化；两者都不提供公开折扣，也不能立即证明账单优惠已生效。</description><source url="https://aws.amazon.com/blogs/machine-learning/manage-private-pricing-offers-for-amazon-bedrock-with-managed-entitlements/">原始来源</source></item><item><title>Orca 介绍如何用更小的结构上下文改进 Discovery AI 查询</title><link>https://aisubhub.top/zh/news/orca-discovery-ai-schema-aware-security-queries/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/orca-discovery-ai-schema-aware-security-queries/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>10 月 13 日案例结合规则匹配、样例检索和验证；900 多个“模型”指云资源结构，而不是大语言模型。</description><source url="https://aws.amazon.com/blogs/machine-learning/natural-language-security-queries-at-orca-security-using-amazon-bedrock/">原始来源</source></item><item><title>Gemini 4 Argon 开放多种入口，订阅与 API 权益需分清</title><link>https://aisubhub.top/zh/news/gemini-4-argon-access-channels/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/gemini-4-argon-access-channels/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>Google 10月13日公告分别列出 Pro、Ultra 应用入口及付费 API，并注明 API 促销期限。</description><source url="https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/">原始来源</source></item><item><title>ShipStation 的 Deckhand 案例区分回答 IT 问题与执行操作</title><link>https://aisubhub.top/zh/news/shipstation-deckhand-agentcore-internal-helpdesk/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/shipstation-deckhand-agentcore-internal-helpdesk/</guid><pubDate>Tue, 20 Oct 2026 10:29:03 GMT</pubDate><description>10 月 13 日案例介绍带确认建单、知识筛选和反馈的 Slack 助手；97.5% 指回复率，不能写成全部问题解决率。</description><source url="https://aws.amazon.com/blogs/machine-learning/how-shipstation-global-built-an-ai-it-help-desk-on-amazon-bedrock-agentcore/">原始来源</source></item><item><title>Anthropic 为何给海量 Agent 代码构建 Scion 版本控制系统</title><link>https://aisubhub.top/zh/news/anthropic-scion-agent-scale-source-control-2026-10-13/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/anthropic-scion-agent-scale-source-control-2026-10-13/</guid><pubDate>Tue, 20 Oct 2026 09:36:37 GMT</pubDate><description>Scion 把历史查询和提交计算放在服务端，按需获取不可变对象；这是一份特殊工作负载的工程案例，不是面向所有团队的 Git 替代方案。</description><source url="https://www.anthropic.com/engineering/redesigning-source-control">原始来源</source></item><item><title>Claude Haiku 5.5 主打高频任务，API 价格按输入长度分档</title><link>https://aisubhub.top/zh/news/claude-haiku-5-5-api-pricing/</link><guid isPermaLink="true">https://aisubhub.top/zh/news/claude-haiku-5-5-api-pricing/</guid><pubDate>Tue, 20 Oct 2026 06:34:05 GMT</pubDate><description>Anthropic 10月7日公告以10万 token 提示长度为门槛，列出不同 API 单价。</description><source url="https://www.anthropic.com/claude-haiku-5-5">原始来源</source></item></channel></rss>