# 端侧推理 · 日跟踪 · 2026-10-06 > 日期:2026-10-06(Asia/Shanghai,周二)。专题正文:`intel/topics/edge-inference.md`。原料:`/workspace/intel/archive/2026-10-06/`(窗口 10-05 早报后→10-06 早);对照 2026-10-05 日跟踪。抓取:主跑 Saved 267 / skipped 28 / errors 40(+ Scale 源修复补抓 14);blocked_hosts:blocksandfiles.com、docs.redhat.com、e.huawei.com、support.huawei.com、www.nextplatform.com、www.reddit.com、www.theregister.com。 ## 摘要 - **本日无端侧推理硬件首发。** - **厂商博客 · RH「MLOps for edge AI」系列 1/4**:核心训练/注册/舰队控制面 vs 边缘推理;"推理不得依赖互联网连接";模型四种打包(应用镜像 / **Modelcar** / 外部存储 / **嵌入 bootc OS 镜像=Red Hat Device Edge,内核+驱动+运行时+模型原子回滚**);异构舰队按 GPU 架构分别编译 TensorRT;硬件性能门禁须在真机跑(Jumpstarter)——**方法论博客,MicroShift 上 KServe 未给版本**。 - 媒体快讯 · Google Project Suncatcher 卫星 4×TPU 跑 Gemma(热约束每次 15 分钟)——太空 AI 实验,轻。 - 续档 · DGX Spark 64GB(10-23 供货)/128GB 涨价。 ## 原文摘录 ### Red Hat · MLOps for edge AI(系列 1/4) - 归档:`redhat/mlops-for-edge-ai-preparing-ai-models-for-deployment-at-the-edge-5f307e38.txt` - URL:https://www.redhat.com/en/blog/mlops-edge-ai-preparing-ai-models-deployment-edge > Additionally, regarding connectivity, 1 important requirement applies to most edge computing use cases: Internet connectivity must not be required for inference to function. > … when the model is combined with a bootc image mode operating system (that is, Red Hat Device Edge), the model can be embedded … in the OS image itself, making the kernel, drivers, runtime, and model a single atomic unit that deploys together and rolls back together.