{"id":132410,"date":"2026-09-04T11:24:26","date_gmt":"2026-09-04T11:24:26","guid":{"rendered":"https:\/\/www.seeedstudio.com\/blog\/?p=132410"},"modified":"2026-09-04T12:05:34","modified_gmt":"2026-09-04T12:05:34","slug":"jetpack-7-2-guide","status":"publish","type":"post","link":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/","title":{"rendered":"JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson"},"content":{"rendered":"\n<p><strong>A practical guide to <\/strong><strong>JetPack<\/strong><strong> 7.2, from the new software foundation and agentic AI workflows to memory <\/strong><strong>optimization<\/strong><strong>, <\/strong><strong>LLM<\/strong><strong> inference, DeepStream, robotics, and production deployment.<\/strong><\/p>\n\n\n\n<p><strong>Version note:<\/strong> NVIDIA\u2019s current JetPack release is JetPack 7.2.1 with Jetson Linux 39.2.1. This article focuses specifically on <strong>JetPack 7.2 with Jetson Linux 39.2<\/strong>, including the software baseline, benchmarks, migration guidance, and Seeed Studio resources built around that release.<\/p>\n\n\n\n<p><a href=\"https:\/\/developer.nvidia.com\/embedded\/jetpack\/downloads\/archive-7.2?utm_source=chatgpt.com\">Download NVIDIA JetPack 7.2<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack_7_2_resource_hub\/?utm_source=chatgpt.com\">Explore the Seeed Studio JetPack 7.2 Resource Hub<\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Introduction<\/h2>\n\n\n\n<p>A JetPack release used to be relatively easy to summarize: a newer Linux base, newer CUDA, updated libraries, and better hardware support.<\/p>\n\n\n\n<p>JetPack 7.2 is more interesting than that.<\/p>\n\n\n\n<p>The release moves NVIDIA Jetson Orin and Jetson Thor onto a common software foundation built around <strong>Jetson Linux 39.2, Ubuntu 24.04, <\/strong><strong>Linux kernel<\/strong><strong> 6.8, and <\/strong><strong>CUDA<\/strong><strong> 13.2.1<\/strong>. It also brings new workflows for agentic AI, memory optimization, production Linux customization, and multi-workload GPU execution.<\/p>\n\n\n\n<p>For developers, the practical question is therefore not simply:<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cWhat new features are included in JetPack 7.2?\u201d<\/p>\n<\/blockquote>\n\n\n\n<p>It is:<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p><strong>\u201cWhat changes in my development and deployment <\/strong><strong>workflow<\/strong><strong> if I move from <\/strong><strong>JetPack<\/strong><strong> 6.x to JetPack 7.2?\u201d<\/strong><\/p>\n<\/blockquote>\n\n\n\n<p>That question matters because the upgrade crosses several software boundaries at once:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Jetson Linux moves from the 36.x generation to 39.2.<\/li>\n\n\n\n<li>Ubuntu moves from 22.04-era JetPack 6.x environments to Ubuntu 24.04.<\/li>\n\n\n\n<li>The Linux kernel moves to 6.8.<\/li>\n\n\n\n<li>CUDA moves to the 13.x generation.<\/li>\n\n\n\n<li>TensorRT and DeepStream move to newer runtime baselines.<\/li>\n\n\n\n<li>Existing drivers, CUDA applications, TensorRT engines, plugins, and system integrations may need to be rebuilt.<\/li>\n<\/ul>\n\n\n\n<p>At the same time, the upgrade opens the door to new capabilities such as NVIDIA Jetson Agent Skills, NemoClaw, official Yocto support, <code>MAXN_SUPER<\/code> on supported Jetson AGX Orin 32GB configurations, and MIG technology preview on Jetson Thor T5000.<\/p>\n\n\n\n<p>This guide looks at JetPack 7.2 from that practical perspective.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img fetchpriority=\"high\" decoding=\"async\" width=\"1030\" height=\"580\" src=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-1030x580.png\" alt=\"\" class=\"wp-image-132434\" srcset=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-1030x580.png 1030w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-300x169.png 300w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-768x432.png 768w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-1536x864.png 1536w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-32x18.png 32w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-1024x576.png 1024w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1.png 1672w\" sizes=\"(max-width: 1030px) 100vw, 1030px\" \/><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">1. JetPack 7.2 at a Glance<\/h2>\n\n\n\n<p>JetPack 7.2 is built around the following software baseline:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><tbody><tr><td>Component<\/td><td>JetPack 7.2<\/td><\/tr><tr><td>Jetson Linux<\/td><td>39.2<\/td><\/tr><tr><td>Kernel<\/td><td>Linux 6.8<\/td><\/tr><tr><td>Root filesystem<\/td><td>Ubuntu 24.04<\/td><\/tr><tr><td>CUDA<\/td><td>13.2.1<\/td><\/tr><tr><td>cuDNN<\/td><td>9.20.0<\/td><\/tr><tr><td>TensorRT<\/td><td>10.16.2<\/td><\/tr><tr><td>DeepStream<\/td><td>9.1<\/td><\/tr><tr><td>Vulkan<\/td><td>1.4<\/td><\/tr><tr><td>VPI<\/td><td>4.1.3<\/td><\/tr><tr><td>PVA<\/td><td>2.9.1<\/td><\/tr><tr><td>Holoscan<\/td><td>3.9.0<\/td><\/tr><tr><td>NVIDIA Container Toolkit<\/td><td>1.19<\/td><\/tr><tr><td>Supported platforms<\/td><td>Jetson Orin family and Jetson Thor<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p>These versions are taken from NVIDIA&#8217;s official JetPack 7.2 archive.<\/p>\n\n\n\n<p><a href=\"https:\/\/developer.nvidia.com\/embedded\/jetpack\/downloads\/archive-7.2?utm_source=chatgpt.com\">NVIDIA JetPack 7.2 Release Page<\/a><\/p>\n\n\n\n<p>The most important point is that JetPack 7.2 is not simply a collection of library updates. It establishes a newer common software foundation across Jetson Orin and Thor.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">2. JetPack 6.2 vs. JetPack 7.2: What Actually Changed?<\/h2>\n\n\n\n<p>At a high level:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><tbody><tr><td>Area<\/td><td>JetPack 6.2<\/td><td>JetPack 7.2<\/td><\/tr><tr><td>Jetson Linux<\/td><td>36.4.3<\/td><td>39.2<\/td><\/tr><tr><td>Ubuntu<\/td><td>22.04<\/td><td>24.04<\/td><\/tr><tr><td>Kernel<\/td><td>5.15 generation<\/td><td>6.8<\/td><\/tr><tr><td>CUDA<\/td><td>12.6<\/td><td>13.2.1<\/td><\/tr><tr><td>TensorRT<\/td><td>10.x generation<\/td><td>10.16.2<\/td><\/tr><tr><td>Jetson platform scope<\/td><td>Orin<\/td><td>Orin + Thor<\/td><\/tr><tr><td>Agentic AI<\/td><td>Early ecosystem<\/td><td>NemoClaw + Jetson Agent Skills<\/td><\/tr><tr><td>Production Linux<\/td><td>Existing BSP workflows<\/td><td>Official Yocto support<\/td><\/tr><tr><td>Orin performance<\/td><td>Standard modes<\/td><td>MAXN_SUPER on supported AGX Orin 32GB<\/td><\/tr><tr><td>Thor GPU partitioning<\/td><td>\u2014<\/td><td>MIG technology preview<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p>The significance is less about any single version number and more about the resulting software lifecycle.<\/p>\n\n\n\n<p>A JetPack 6.x application that depends on custom kernel modules, device trees, CUDA binaries, TensorRT engines, camera drivers, or GStreamer plugins should be treated as a migration project rather than a routine package upgrade. Seeed&#8217;s JetPack 7.2 documentation explicitly recommends rebuilding these components against the new software baseline.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">3. JetPack 7.2 Is a Platform Migration, Not Just an Update<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" width=\"1030\" height=\"773\" src=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_2-1-1030x773.png\" alt=\"\" class=\"wp-image-132438\" srcset=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_2-1-1030x773.png 1030w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_2-1-300x225.png 300w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_2-1-768x576.png 768w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_2-1-32x24.png 32w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_2-1-1024x768.png 1024w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_2-1.png 1400w\" sizes=\"(max-width: 1030px) 100vw, 1030px\" \/><\/figure>\n\n\n\n<p>The most important migration changes are underneath the application.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Ubuntu 24.04<\/h3>\n\n\n\n<p>Moving to Ubuntu 24.04 means reviewing:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>system packages;<\/li>\n\n\n\n<li>repositories;<\/li>\n\n\n\n<li>Python environments;<\/li>\n\n\n\n<li>system services;<\/li>\n\n\n\n<li>container dependencies;<\/li>\n\n\n\n<li>build scripts;<\/li>\n\n\n\n<li>installation instructions.<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Linux 6.8<\/h3>\n\n\n\n<p>A newer kernel affects custom hardware integration.<\/p>\n\n\n\n<p>Projects using:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Wi-Fi modules;<\/li>\n\n\n\n<li>CSI or GMSL cameras;<\/li>\n\n\n\n<li>CAN;<\/li>\n\n\n\n<li>Ethernet;<\/li>\n\n\n\n<li>PCIe;<\/li>\n\n\n\n<li>USB;<\/li>\n\n\n\n<li>custom carrier boards;<\/li>\n\n\n\n<li>out-of-tree kernel modules<\/li>\n<\/ul>\n\n\n\n<p>should rebuild and validate those components against Jetson Linux 39.2.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">CUDA 13.x<\/h3>\n\n\n\n<p>CUDA applications compiled against the previous JetPack environment should not simply be copied to the new system.<\/p>\n\n\n\n<p>Rebuild them against the JetPack 7.2 toolchain and validate:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>CUDA compatibility;<\/li>\n\n\n\n<li>dependencies;<\/li>\n\n\n\n<li>compiler configuration;<\/li>\n\n\n\n<li>architecture flags;<\/li>\n\n\n\n<li>runtime behavior.<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">TensorRT engines<\/h3>\n\n\n\n<p>Serialized TensorRT engines are particularly important.<\/p>\n\n\n\n<p><strong>Do not assume a TensorRT engine generated under <\/strong><strong>JetPack<\/strong><strong> 6.x can be reused under JetPack 7.2.<\/strong><\/p>\n\n\n\n<p>For JetPack 7.2 deployments, rebuild the engine against the target TensorRT stack.<\/p>\n\n\n\n<p>This same principle applies to custom TensorRT plugins and other binary components.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack72_deep_dive\/?utm_source=chatgpt.com\">Seeed JetPack 7.2 Deep Dive<\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">4. The Agentic AI Layer: Jetson Agent Skills<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" width=\"1030\" height=\"917\" src=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_3-1030x917.png\" alt=\"\" class=\"wp-image-132437\" srcset=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_3-1030x917.png 1030w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_3-300x267.png 300w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_3-768x684.png 768w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_3-32x28.png 32w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_3-1024x912.png 1024w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_3.png 1329w\" sizes=\"(max-width: 1030px) 100vw, 1030px\" \/><\/figure>\n\n\n\n<p>One of the most distinctive additions in JetPack 7.2 is the introduction of NVIDIA&#8217;s <strong>Jetson Agent Skills<\/strong>.<\/p>\n\n\n\n<p>The idea is straightforward: instead of asking a general-purpose coding agent to understand every Jetson-specific configuration detail from scratch, Jetson-specific knowledge and procedures can be packaged into reusable workflows.<\/p>\n\n\n\n<p>NVIDIA describes three major categories:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Jetson Linux customization;<\/li>\n\n\n\n<li>memory optimization;<\/li>\n\n\n\n<li>model benchmarking and diagnostics.<\/li>\n<\/ul>\n\n\n\n<p>There are also workflows around DeepStream and vision applications.<\/p>\n\n\n\n<p>This creates an interesting shift in the development workflow.<\/p>\n\n\n\n<p>Previously:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Documentation\n    \u2193\nDeveloper\n    \u2193\nManual configuration\n    \u2193\nBuild\n    \u2193\nFlash\n    \u2193\nTest\n    \u2193\nDebug<\/code><\/pre>\n\n\n\n<p>With agent-assisted workflows:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Developer intent\n       \u2193\nCoding Agent\n       \u2193\nJetson-specific Skills\n       \u2193\nConfiguration \/ Build \/ Benchmark\n       \u2193\nValidation<\/code><\/pre>\n\n\n\n<p>The important word here is <strong>workflow<\/strong>.<\/p>\n\n\n\n<p>Agent Skills do not magically make hardware development autonomous. They provide structured, repeatable instructions that an agent can execute against Jetson-specific tasks.<\/p>\n\n\n\n<p><a href=\"https:\/\/github.com\/NVIDIA-AI-IOT\/jetson-device-skills?utm_source=chatgpt.com\">NVIDIA Jetson Device Skills on GitHub<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/github.com\/NVIDIA-AI-IOT\/jetson-bsp-skills?utm_source=chatgpt.com\">NVIDIA Jetson BSP Skills on GitHub<\/a><\/p>\n\n\n\n<p>For Seeed developers, this is particularly relevant because Jetson deployment often involves hardware-specific details that general coding agents do not automatically know.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/rapid_prototyping_on_jetson_with_nvidia_skills\/?utm_source=chatgpt.com\">Seeed: Rapid Prototyping on Jetson with NVIDIA Skills<\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">5. NemoClaw: Bringing Agentic Workflows to the Edge<\/h2>\n\n\n\n<p>JetPack 7.2 also provides native support for installing NVIDIA NemoClaw with a single command.<\/p>\n\n\n\n<p>NemoClaw is an open-source stack designed to add privacy and security controls around OpenClaw-based assistants and agentic workflows. NVIDIA positions it as a way to bring agentic AI into robotics, industrial automation, vision, and other edge applications.<\/p>\n\n\n\n<p>This becomes especially interesting when an agent can move beyond answering questions and interact with physical systems.<\/p>\n\n\n\n<p>A simplified architecture looks like:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>User\n  \u2193\nAgent\n  \u2193\nReasoning \/ Planning\n  \u2193\nTools\n  \u2193\nJetson\n  \u2193\nVision \/ Sensors\n  \u2193\nRobot \/ Actuator<\/code><\/pre>\n\n\n\n<p>Seeed has already demonstrated this direction with a Jetson Thor + reBot Arm B601 workflow.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/control_rebot_arm_with_nemoclaw_on_nvidia_jetson_thor\/?utm_source=chatgpt.com\">Seeed: Control reBot Arm B601 with NemoClaw on Jetson Thor<\/a><\/p>\n\n\n\n<p>The important takeaway is not that every robot should immediately be controlled by an autonomous agent.<\/p>\n\n\n\n<p>It is that JetPack 7.2 makes the software path from <strong>AI reasoning \u2192 tools \u2192 perception \u2192 physical execution<\/strong> substantially easier to prototype.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">6. The Hidden Upgrade: Memory Efficiency<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1030\" height=\"913\" src=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_4-1030x913.png\" alt=\"\" class=\"wp-image-132440\" srcset=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_4-1030x913.png 1030w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_4-300x266.png 300w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_4-768x681.png 768w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_4-32x28.png 32w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_4-1024x908.png 1024w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_4.png 1280w\" sizes=\"(max-width: 1030px) 100vw, 1030px\" \/><\/figure>\n\n\n\n<p>For edge AI, memory can be more important than raw compute.<\/p>\n\n\n\n<p>A Jetson system does not have a separate pool of memory for every application layer. The available DRAM is shared across:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>operating system services;<\/li>\n\n\n\n<li>GPU workloads;<\/li>\n\n\n\n<li>camera and multimedia components;<\/li>\n\n\n\n<li>model weights;<\/li>\n\n\n\n<li>TensorRT engines;<\/li>\n\n\n\n<li>runtime workspaces;<\/li>\n\n\n\n<li>KV cache;<\/li>\n\n\n\n<li>robotics middleware;<\/li>\n\n\n\n<li>application services.<\/li>\n<\/ul>\n\n\n\n<p>That means reducing system overhead can directly increase the memory available for AI workloads.<\/p>\n\n\n\n<p>Seeed&#8217;s JetPack 7.2 memory research highlights this point: JetPack 7.2 does <strong>not<\/strong> add physical DRAM or automatically shrink every model. The improvement comes from the software baseline and the ability to systematically optimize how memory is allocated.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack_7_2_memory_optimization\/?utm_source=chatgpt.com\">Seeed JetPack 7.2 Memory Optimization<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack_7_2_memory_optimization_deep_dive\/?utm_source=chatgpt.com\">Seeed JetPack 7.2 Memory Optimization Deep Dive<\/a><\/p>\n\n\n\n<p>A useful way to think about the problem is:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Physical DRAM\n\u2502\n\u251c\u2500\u2500 OS + services\n\u251c\u2500\u2500 GPU \/ CUDA runtime\n\u251c\u2500\u2500 Camera \/ multimedia\n\u251c\u2500\u2500 Model weights\n\u251c\u2500\u2500 TensorRT workspace\n\u251c\u2500\u2500 KV cache\n\u251c\u2500\u2500 Robotics middleware\n\u2514\u2500\u2500 Application memory<\/code><\/pre>\n\n\n\n<p>Every megabyte removed from unnecessary overhead becomes potential headroom for the workload that actually matters.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">7. Seeed&#8217;s JetPack 6.2 vs. JetPack 7.2 LLM Test<\/h2>\n\n\n\n<p>Seeed compared JetPack 6.2 and JetPack 7.2 using a Qwen3.5-27B Q4_K_M workload on a 32GB-class Jetson AGX Orin configuration.<\/p>\n\n\n\n<p>The published results were:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><tbody><tr><td>Metric<\/td><td>JetPack 6.2<\/td><td>JetPack 7.2<\/td><td>Observed change<\/td><\/tr><tr><td>Memory after model load<\/td><td>24.6 GB \/ 30 GB<\/td><td>14.7 GB \/ 30 GB<\/td><td>~40% lower<\/td><\/tr><tr><td>GPU frequency during inference<\/td><td>930 MHz<\/td><td>1.36 GHz<\/td><td>Higher boost frequency<\/td><\/tr><tr><td>Prompt processing<\/td><td>18.2 tokens\/s<\/td><td>25.8 tokens\/s<\/td><td>~41.8% higher<\/td><\/tr><tr><td>Token generation<\/td><td>4.3 tokens\/s<\/td><td>5.5 tokens\/s<\/td><td>~27.9% higher<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p>The most practical result is memory headroom.<\/p>\n\n\n\n<p>In this test configuration, the JetPack 7.2 run retained roughly <strong>10 <\/strong><strong>GB<\/strong><strong> more memory headroom<\/strong> after model loading. That matters when an LLM has to coexist with camera processing, robotics middleware, databases, APIs, or other edge services.<\/p>\n\n\n\n<p>These numbers should be interpreted as results from Seeed&#8217;s published test configuration, not as universal performance guarantees for every Jetson device or workload.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack72_deep_dive\/?utm_source=chatgpt.com\">Seeed JetPack 7.2 Deep Dive &amp; Benchmark<\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">8. Think About LLM Memory as a Budget<\/h2>\n\n\n\n<p>A useful LLM deployment model is:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Available DRAM\n    -\nSystem footprint\n    -\nInference runtime\n    -\nModel weights\n    -\nKV cache\n    -\nApplication services\n    =\nActual headroom<\/code><\/pre>\n\n\n\n<p>This is why a smaller system footprint can have an outsized impact on an edge LLM application.<\/p>\n\n\n\n<p>Seeed&#8217;s Orin Nano 8GB observations found a historical JetPack 6.2 state using roughly 1.4 GiB at boot compared with slightly more than 800 MiB in one JetPack 7.2 configuration\u2014around 600 MiB of difference in that particular image and service configuration.<\/p>\n\n\n\n<p>That does <strong>not<\/strong> mean every JetPack 7.2 system will automatically save exactly 600 MiB.<\/p>\n\n\n\n<p>Desktop services, containers, camera paths, carrier-board BSP settings, and measurement timing all affect the result.<\/p>\n\n\n\n<p>The correct workflow is:<\/p>\n\n\n\n<p><strong>Measure \u2192 identify \u2192 optimize \u2192 measure again.<\/strong><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">9. TensorRT Edge-LLM: Rebuilding the Inference Stack<\/h2>\n\n\n\n<p>JetPack 7.2 also creates a newer baseline for optimized LLM inference.<\/p>\n\n\n\n<p>Seeed&#8217;s TensorRT Edge-LLM guide documents JetPack 7.2 workflows for Jetson Orin and Thor, including engine generation and benchmarking. For the documented Orin path, the workflow uses CUDA 13.2 and supports FP16, INT8, and INT4 runtime paths.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/deploy_tensorrt_edge_llm_on_jetpack7.2\/?utm_source=chatgpt.com\">Seeed: Deploy TensorRT Edge-LLM on JetPack 7.2<\/a><\/p>\n\n\n\n<p>One migration rule is particularly important:<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p><strong>Rebuild your inference engines for <\/strong><strong>JetPack<\/strong><strong> 7.2.<\/strong><\/p>\n<\/blockquote>\n\n\n\n<p>Do not treat serialized engines as portable application files across JetPack generations.<\/p>\n\n\n\n<p>A typical workflow becomes:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Model\n  \u2193\nExport\n  \u2193\nJetPack 7.2-compatible build\n  \u2193\nTensorRT Edge-LLM\n  \u2193\nEngine generation\n  \u2193\nTarget Jetson\n  \u2193\nBenchmark<\/code><\/pre>\n\n\n\n<p>For production work, benchmark more than tokens per second.<\/p>\n\n\n\n<p>Track:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>engine build memory;<\/li>\n\n\n\n<li>engine load memory;<\/li>\n\n\n\n<li>time to first token;<\/li>\n\n\n\n<li>prompt processing throughput;<\/li>\n\n\n\n<li>decode throughput;<\/li>\n\n\n\n<li>GPU frequency;<\/li>\n\n\n\n<li>power mode;<\/li>\n\n\n\n<li>temperature;<\/li>\n\n\n\n<li>board power.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">10. DeepStream 9.1: A New Video Analytics Baseline<\/h2>\n\n\n\n<p>JetPack 7.2 also changes the baseline for video analytics.<\/p>\n\n\n\n<p>For JetPack 7.2, <strong>DeepStream 9.1<\/strong> is the relevant Jetson release. Seeed&#8217;s documentation identifies DeepStream 9.1 as the first DeepStream release whose Jetson platform table explicitly targets JetPack 7.2 GA \/ Jetson Linux 39.2.<\/p>\n\n\n\n<p><a href=\"https:\/\/github.com\/Seeed-Studio\/wiki-documents\/blob\/docusaurus-version\/sites\/en\/docs\/Edge\/NVIDIA_Jetson\/JetPack_7_2\/AI_Inference\/JetPack_7_2_DeepStream.md?utm_source=chatgpt.com\">Seeed: DeepStream on JetPack 7.2<\/a><\/p>\n\n\n\n<p>The baseline includes:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>DeepStream 9.1;<\/li>\n\n\n\n<li>Jetson Linux 39.2;<\/li>\n\n\n\n<li>CUDA 13.2;<\/li>\n\n\n\n<li>TensorRT 10.16.1.7 within the DeepStream 9.1 package baseline;<\/li>\n\n\n\n<li>cuDNN 9.20.0.46;<\/li>\n\n\n\n<li>GStreamer 1.24.2;<\/li>\n\n\n\n<li>OpenCV 4.8.0.<\/li>\n<\/ul>\n\n\n\n<p>Notice that the TensorRT version here is the <strong>DeepStream 9.1 package baseline<\/strong>, rather than the top-level JetPack 7.2 TensorRT package version. Keeping these version contexts separate avoids confusing the two software baselines.<\/p>\n\n\n\n<p>DeepStream 9.1 also introduces a stronger agent-assisted development workflow, including coding-agent capabilities and agentic skills for pipeline construction, debugging, multi-camera tracking, calibration, and inference configuration.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">11. From Development Image to Production Image: Yocto<\/h2>\n\n\n\n<p>For a prototype, Ubuntu-based JetPack is often the fastest way to get started.<\/p>\n\n\n\n<p>For a production device, the requirements can be different.<\/p>\n\n\n\n<p>You may want:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>a smaller system image;<\/li>\n\n\n\n<li>controlled package versions;<\/li>\n\n\n\n<li>reproducible builds;<\/li>\n\n\n\n<li>a fixed boot configuration;<\/li>\n\n\n\n<li>custom services;<\/li>\n\n\n\n<li>predictable update behavior;<\/li>\n\n\n\n<li>a defined hardware configuration.<\/li>\n<\/ul>\n\n\n\n<p>JetPack 7.2 adds official Yocto Project support, providing another route to custom Linux distributions for Jetson. NVIDIA specifically positions this capability around customized production systems and software efficiency.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/build_and_flash_yocto_for_seeed_jetson_carrier_boards\/?utm_source=chatgpt.com\">Seeed: Build and Flash a Yocto Image for Seeed Jetson Carrier Boards<\/a><\/p>\n\n\n\n<p>For Seeed hardware, the practical workflow is:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Hardware configuration\n       \u2193\nYocto metadata\n       \u2193\nBSP + kernel + drivers\n       \u2193\nCustom root filesystem\n       \u2193\nProduction image\n       \u2193\nFlash\n       \u2193\nValidation<\/code><\/pre>\n\n\n\n<p>Yocto is not necessarily the best choice for every prototype.<\/p>\n\n\n\n<p>It becomes increasingly valuable when you need to control the entire software image and maintain it over a fleet.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">12. MAXN_SUPER: More Performance from Supported AGX Orin 32GB Configurations<\/h2>\n\n\n\n<p>JetPack 7.2 introduces <code>MAXN_SUPER<\/code> for supported Jetson AGX Orin 32GB configurations.<\/p>\n\n\n\n<p>NVIDIA reports an increase in headline AI performance from <strong>200 <\/strong><strong>TOPS<\/strong><strong> to 241 TOPS<\/strong>, alongside higher GPU frequencies and a higher power envelope.<\/p>\n\n\n\n<p><a href=\"https:\/\/developer.nvidia.com\/blog\/deploy-agentic-ready-ai-at-the-edge-with-memory-efficiency-in-nvidia-jetpack-7-2?utm_source=chatgpt.com\">NVIDIA: JetPack 7.2 Technical Overview<\/a><\/p>\n\n\n\n<p>This is an example of why JetPack updates can change the value of existing hardware.<\/p>\n\n\n\n<p>The hardware itself has not suddenly gained more physical compute units. Instead, the newer software and power configuration can expose additional performance within supported operating conditions.<\/p>\n\n\n\n<p>Actual application performance will still depend on:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>workload;<\/li>\n\n\n\n<li>model;<\/li>\n\n\n\n<li>precision;<\/li>\n\n\n\n<li>power mode;<\/li>\n\n\n\n<li>thermal conditions;<\/li>\n\n\n\n<li>memory behavior;<\/li>\n\n\n\n<li>software optimization.<\/li>\n<\/ul>\n\n\n\n<p>So the useful question is not:<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cDoes MAXN_SUPER make every application 20% faster?\u201d<\/p>\n<\/blockquote>\n\n\n\n<p>It is:<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p><strong>\u201cDoes my workload benefit enough from the higher performance envelope to justify the additional power and thermal requirements?\u201d<\/strong><\/p>\n<\/blockquote>\n\n\n\n<h2 class=\"wp-block-heading\">13. MIG on Jetson Thor<\/h2>\n\n\n\n<p>JetPack 7.2 introduces <strong>Multi-Instance GPU<\/strong><strong> (<\/strong><strong>MIG<\/strong><strong>) support on Jetson Thor T5000 as a technology preview<\/strong>.<\/p>\n\n\n\n<p>MIG allows the GPU to be partitioned into isolated instances so that different workloads can share the same SoC with more predictable resource allocation. NVIDIA describes this as particularly relevant to mixed-criticality workloads in robotics, industrial automation, and physical AI.<\/p>\n\n\n\n<p>A conceptual architecture looks like:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Jetson Thor\n\u2502\n\u251c\u2500\u2500 MIG Instance A\n\u2502     \u251c\u2500\u2500 Perception\n\u2502     \u2514\u2500\u2500 Control\n\u2502\n\u2514\u2500\u2500 MIG Instance B\n      \u251c\u2500\u2500 Generative AI\n      \u2514\u2500\u2500 Visualization<\/code><\/pre>\n\n\n\n<p>This is especially interesting for robots because modern systems increasingly combine:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>real-time control;<\/li>\n\n\n\n<li>perception;<\/li>\n\n\n\n<li>sensor fusion;<\/li>\n\n\n\n<li>planning;<\/li>\n\n\n\n<li>generative AI;<\/li>\n\n\n\n<li>safety monitoring.<\/li>\n<\/ul>\n\n\n\n<p>The value of GPU partitioning is therefore not simply \u201cmore performance.\u201d<\/p>\n\n\n\n<p>It is <strong>more predictable sharing of compute resources<\/strong>.<\/p>\n\n\n\n<p>Because MIG on Jetson Thor is a technology preview in JetPack 7.2, production designs should validate the exact supported configuration and workload behavior before relying on it.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">14. Camera and Multimedia: An Area to Validate Carefully<\/h2>\n\n\n\n<p>Camera support is one of the areas where a JetPack migration should never be treated as a simple software replacement.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1030\" height=\"579\" src=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_5-1030x579.png\" alt=\"\" class=\"wp-image-132441\" srcset=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_5-1030x579.png 1030w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_5-300x169.png 300w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_5-768x432.png 768w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_5-32x18.png 32w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_5-1024x576.png 1024w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_5.png 1280w\" sizes=\"(max-width: 1030px) 100vw, 1030px\" \/><\/figure>\n\n\n\n<p>A Jetson application may depend on:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>CSI;<\/li>\n\n\n\n<li>GMSL;<\/li>\n\n\n\n<li>Argus;<\/li>\n\n\n\n<li>V4L2;<\/li>\n\n\n\n<li>GStreamer;<\/li>\n\n\n\n<li>ISP configuration;<\/li>\n\n\n\n<li>device trees;<\/li>\n\n\n\n<li>sensor drivers;<\/li>\n\n\n\n<li>camera synchronization;<\/li>\n\n\n\n<li>hardware codecs.<\/li>\n<\/ul>\n\n\n\n<p>With the move to Jetson Linux 39.2 and Linux kernel 6.8, these components should be explicitly revalidated.<\/p>\n\n\n\n<p>For a production camera system, test:<\/p>\n\n\n\n<ol start=\"1\" class=\"wp-block-list\">\n<li>camera enumeration;<\/li>\n\n\n\n<li>sensor initialization;<\/li>\n\n\n\n<li>resolution and frame rate;<\/li>\n\n\n\n<li>exposure and ISP behavior;<\/li>\n\n\n\n<li>multi-camera synchronization;<\/li>\n\n\n\n<li>hardware encoding\/decoding;<\/li>\n\n\n\n<li>zero-copy paths;<\/li>\n\n\n\n<li>long-duration stability.<\/li>\n<\/ol>\n\n\n\n<p>Do not assume that a camera pipeline working under JetPack 6.x will behave identically after migration.<\/p>\n\n\n\n<p>Seeed&#8217;s JetPack 7.2 Resource Hub collects the current camera, wireless, and multimedia migration material as it becomes available.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack_7_2_resource_hub\/?utm_source=chatgpt.com\">Seeed JetPack 7.2 Resource Hub<\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">16. Real-World Application: Industrial Vision<\/h2>\n\n\n\n<p>One useful way to evaluate a platform release is to stop looking at individual SDK features and build an actual application.<\/p>\n\n\n\n<p>Seeed&#8217;s Industrial Vision Monitoring workflow demonstrates a JetPack 7.2 application using:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>USB camera input;<\/li>\n\n\n\n<li>YOLO-based detection;<\/li>\n\n\n\n<li>PPE detection;<\/li>\n\n\n\n<li>VLM-based scene understanding;<\/li>\n\n\n\n<li>browser-based visualization.<\/li>\n<\/ul>\n\n\n\n<p>The workflow has been validated on Seeed industrial Jetson platforms running JetPack 7.2 \/ L4T 39.2.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Demo Results<\/h3>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1030\" height=\"478\" src=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_6-1030x478.png\" alt=\"\" class=\"wp-image-132442\" srcset=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_6-1030x478.png 1030w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_6-300x139.png 300w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_6-768x357.png 768w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_6-1536x713.png 1536w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_6-32x15.png 32w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_6-1024x476.png 1024w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_6.png 2026w\" sizes=\"(max-width: 1030px) 100vw, 1030px\" \/><\/figure>\n\n\n\n<p>Idle UI before the camera stream starts.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1030\" height=\"819\" src=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_7-1030x819.png\" alt=\"\" class=\"wp-image-132443\" srcset=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_7-1030x819.png 1030w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_7-300x239.png 300w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_7-768x611.png 768w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_7-1536x1222.png 1536w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_7-32x25.png 32w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_7-1024x815.png 1024w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_7.png 1995w\" sizes=\"(max-width: 1030px) 100vw, 1030px\" \/><\/figure>\n\n\n\n<p>VLM behavior alert \u2014 phone use in the work area.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1030\" height=\"781\" src=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_8-1030x781.png\" alt=\"\" class=\"wp-image-132444\" srcset=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_8-1030x781.png 1030w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_8-300x228.png 300w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_8-768x583.png 768w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_8-32x24.png 32w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_8-1024x777.png 1024w, https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_8.png 1280w\" sizes=\"(max-width: 1030px) 100vw, 1030px\" \/><\/figure>\n\n\n\n<p>YOLO PPE alert \u2014 missing safety helmet.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/industrial_vision_monitoring_on_industrial\/?utm_source=chatgpt.com\">Seeed: Industrial Vision Monitoring on JetPack 7.2<\/a><\/p>\n\n\n\n<p>The architecture illustrates an important trend:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Camera\n  \u2193\nJetson\n  \u2193\nReal-time detection\n  \u2193\nEvent \/ PPE \/ object filtering\n  \u2193\nHigher-level visual reasoning\n  \u2193\nHuman-readable alert<\/code><\/pre>\n\n\n\n<p>This is where the JetPack 7.2 improvements begin to connect.<\/p>\n\n\n\n<p>The value is not just better inference speed.<\/p>\n\n\n\n<p>It is the ability to combine <strong>vision + LLM\/VLM + edge services + hardware interfaces<\/strong> within one deployable system.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">16. Physical AI: GR00T N1.7 on JetPack 7.2<\/h2>\n\n\n\n<p>The same idea extends to robotics.<\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<iframe title=\"(Base on Tensorrt) AGX Orin infer gr00t n1.7\" width=\"640\" height=\"360\" src=\"https:\/\/www.youtube.com\/embed\/H3rQHnB-gaI?feature=oembed\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe>\n<\/div><\/figure>\n\n\n\n<p>Seeed has published a JetPack 7.2 workflow for deploying the full-size <strong>GR00T N1.7<\/strong> model stack on Jetson AGX Orin.<\/p>\n\n\n\n<p>The workflow covers multiple TensorRT engines across the GR00T pipeline, including:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>ViT;<\/li>\n\n\n\n<li>LLM;<\/li>\n\n\n\n<li>vision-language attention;<\/li>\n\n\n\n<li>state encoder;<\/li>\n\n\n\n<li>action encoder;<\/li>\n\n\n\n<li>DiT action expert;<\/li>\n\n\n\n<li>action decoder.<\/li>\n<\/ul>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/cn\/deploy_full_weight_gr00t_n1.7_tensorrt_jetpack7.2_agx_orin\/?utm_source=chatgpt.com\">Seeed: Deploy Full-Weight GR00T N1.7 with TensorRT on JetPack 7.2 and AGX Orin<\/a><\/p>\n\n\n\n<p>For reproducibility, developers should follow the environment and pinned repository state documented by Seeed rather than assuming that every version of PyTorch, Transformers, or other dependencies will remain interchangeable.<\/p>\n\n\n\n<p>This is another example of why JetPack should be considered an <strong>AI deployment foundation<\/strong>, not just an operating-system image.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">17. How Should You Upgrade to JetPack 7.2?<\/h2>\n\n\n\n<p>The safest way to approach JetPack 7.2 is to treat it as a controlled migration.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Step 1: Confirm Hardware Support<\/h3>\n\n\n\n<p>First confirm:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Jetson module;<\/li>\n\n\n\n<li>carrier board;<\/li>\n\n\n\n<li>BSP;<\/li>\n\n\n\n<li>camera configuration;<\/li>\n\n\n\n<li>peripherals;<\/li>\n\n\n\n<li>required drivers.<\/li>\n<\/ul>\n\n\n\n<p>Do not assume that because a module belongs to the Jetson family, every carrier configuration has an identical JetPack 7.2 image.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Step 2: Back Up the System<\/h3>\n\n\n\n<p>Back up:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>application source;<\/li>\n\n\n\n<li>configuration;<\/li>\n\n\n\n<li>containers;<\/li>\n\n\n\n<li>persistent volumes;<\/li>\n\n\n\n<li>calibration data;<\/li>\n\n\n\n<li>device-tree source;<\/li>\n\n\n\n<li>custom kernel modules;<\/li>\n\n\n\n<li>credentials and deployment configuration.<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Step 3: Choose the Installation Strategy<\/h3>\n\n\n\n<p>For a new development system or a major migration, Seeed recommends a full flash as the default path.<\/p>\n\n\n\n<p>Image-based OTA is appropriate only when the exact source and target image, board configuration, partition layout, and rollback procedure have been validated.<\/p>\n\n\n\n<p>A cross-major <code>apt upgrade<\/code> should <strong>not<\/strong> be treated as a JetPack 6.x \u2192 7.2 migration mechanism.<\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/flash_and_ota_jetpack_7.2\/?utm_source=chatgpt.com\">Seeed: Flash and OTA to JetPack 7.2<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetson_developtool_overview\/?utm_source=chatgpt.com\">Seeed Jetson DevelopTool<\/a><\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Step 4: Rebuild Dependencies<\/h3>\n\n\n\n<p>Rebuild:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>kernel modules;<\/li>\n\n\n\n<li>CUDA applications;<\/li>\n\n\n\n<li>TensorRT engines;<\/li>\n\n\n\n<li>TensorRT plugins;<\/li>\n\n\n\n<li>GStreamer plugins;<\/li>\n\n\n\n<li>camera drivers;<\/li>\n\n\n\n<li>custom Python\/C++ extensions.<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Step 5: Validate the Hardware<\/h3>\n\n\n\n<p>Check:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>boot;<\/li>\n\n\n\n<li>Ethernet;<\/li>\n\n\n\n<li>Wi-Fi;<\/li>\n\n\n\n<li>USB;<\/li>\n\n\n\n<li>PCIe;<\/li>\n\n\n\n<li>CAN;<\/li>\n\n\n\n<li>cameras;<\/li>\n\n\n\n<li>storage;<\/li>\n\n\n\n<li>thermal behavior;<\/li>\n\n\n\n<li>power modes.<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Step 6: Benchmark the Actual Application<\/h3>\n\n\n\n<p>Record:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>latency;<\/li>\n\n\n\n<li>throughput;<\/li>\n\n\n\n<li>memory;<\/li>\n\n\n\n<li>GPU frequency;<\/li>\n\n\n\n<li>power;<\/li>\n\n\n\n<li>temperature;<\/li>\n\n\n\n<li>dropped frames;<\/li>\n\n\n\n<li>application stability.<\/li>\n<\/ul>\n\n\n\n<p>Only after this should you optimize the workload further.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">18. Who Should Upgrade to JetPack 7.2?<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">New Jetson Orin projects<\/h3>\n\n\n\n<p><strong>Strong candidate.<\/strong><\/p>\n\n\n\n<p>If your required drivers, BSP, cameras, and applications already support JetPack 7.2, starting on the newer baseline avoids building new systems on an older software generation.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">LLM and generative AI applications<\/h3>\n\n\n\n<p><strong>Strong candidate.<\/strong><\/p>\n\n\n\n<p>The combination of CUDA 13.x, updated TensorRT, memory optimization workflows, Agent Skills, and newer inference stacks makes JetPack 7.2 particularly interesting for memory-constrained generative AI.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Robotics and physical AI<\/h3>\n\n\n\n<p><strong>Strong candidate.<\/strong><\/p>\n\n\n\n<p>NemoClaw, Jetson Agent Skills, GR00T workflows, improved memory efficiency, and the newer platform architecture all align well with robotics workloads.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Existing JetPack 6.x production systems<\/h3>\n\n\n\n<p><strong>Evaluate carefully.<\/strong><\/p>\n\n\n\n<p>If your system is stable and highly customized, the engineering cost of migration may be significant.<\/p>\n\n\n\n<p>Upgrade when you have a clear reason:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>newer AI models;<\/li>\n\n\n\n<li>memory constraints;<\/li>\n\n\n\n<li>new software dependencies;<\/li>\n\n\n\n<li>new Jetson platform support;<\/li>\n\n\n\n<li>new deployment architecture;<\/li>\n\n\n\n<li>long-term software lifecycle requirements.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">19. A Simple JetPack 7.2 Decision Tree<\/h2>\n\n\n\n<pre class=\"wp-block-code\"><code>Do you need a new Jetson project?\n        |\n       Yes\n        \u2193\nDoes your BSP \/ camera \/ driver stack support JP7.2?\n        |\n   \u250c\u2500\u2500\u2500\u2500\u2534\u2500\u2500\u2500\u2500\u2510\n   Yes       No\n    \u2193         \u2193\n Start      Validate\n JP7.2      hardware stack\n    |\n    \u2193\nDo you need LLM \/ robotics \/ vision workloads?\n    |\n   Yes\n    \u2193\nMeasure memory + inference + thermal behavior\n    |\n    \u2193\nOptimize\n    |\n    \u2193\nValidate\n    |\n    \u2193\nDeploy<\/code><\/pre>\n\n\n\n<p>For an existing JetPack 6.x product:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Existing JP6.x product\n        \u2193\nInventory dependencies\n        \u2193\nBackup\n        \u2193\nFlash JP7.2 test device\n        \u2193\nRebuild\n        \u2193\nPeripheral validation\n        \u2193\nAI workload validation\n        \u2193\nLong-duration test\n        \u2193\nProduction decision<\/code><\/pre>\n\n\n\n<h2 class=\"wp-block-heading\">20. <strong>What JetPack 7.2 Really Changes<\/strong><\/h2>\n\n\n\n<p>The easiest way to misunderstand JetPack 7.2 is to reduce it to:<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>Ubuntu 24.04 + CUDA 13 + newer libraries.<\/p>\n<\/blockquote>\n\n\n\n<p>The more useful way to see it is as a shift in the Jetson development stack.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Platform<\/h3>\n\n\n\n<p>Jetson Orin and Thor move toward a common software foundation.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">AI<\/h3>\n\n\n\n<p>The software stack is increasingly designed around generative AI and agentic workflows.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Memory<\/h3>\n\n\n\n<p>System-level memory optimization becomes a first-class deployment concern.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Agents<\/h3>\n\n\n\n<p>Jetson-specific knowledge can be exposed through reusable agent workflows.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Inference<\/h3>\n\n\n\n<p>TensorRT and Edge-LLM workflows provide newer paths for optimized local inference.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Video<\/h3>\n\n\n\n<p>DeepStream 9.1 establishes the relevant JetPack 7.2 video analytics baseline.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Production<\/h3>\n\n\n\n<p>Yocto provides a path toward reproducible, customized Linux images.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Robotics<\/h3>\n\n\n\n<p>NemoClaw, GR00T, vision, middleware, and hardware interfaces can be combined into increasingly capable physical AI systems.<\/p>\n\n\n\n<p>That is why JetPack 7.2 is better understood as a <strong>software foundation for the next generation of edge AI applications<\/strong> rather than simply another SDK update.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">21. <strong>JetPack 7.2 Resources<\/strong><\/h2>\n\n\n\n<p>If you are working with JetPack 7.2, these resources are the best places to start.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">NVIDIA<\/h3>\n\n\n\n<p><a href=\"https:\/\/developer.nvidia.com\/embedded\/jetpack\/downloads\/archive-7.2?utm_source=chatgpt.com\">NVIDIA JetPack 7.2 Download &amp; Archive<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/developer.nvidia.com\/embedded\/jetpack-archive?utm_source=chatgpt.com\">NVIDIA JetPack Archive<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/developer.nvidia.com\/blog\/deploy-agentic-ready-ai-at-the-edge-with-memory-efficiency-in-nvidia-jetpack-7-2?utm_source=chatgpt.com\">NVIDIA Technical Blog: JetPack 7.2<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/github.com\/NVIDIA-AI-IOT\/jetson-device-skills?utm_source=chatgpt.com\">NVIDIA Jetson Device Skills<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/github.com\/NVIDIA-AI-IOT\/jetson-bsp-skills?utm_source=chatgpt.com\">NVIDIA Jetson BSP Skills<\/a><\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Seeed Studio<\/h3>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack_7_2_resource_hub\/?utm_source=chatgpt.com\">Seeed JetPack 7.2 Resource Hub<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack72_deep_dive\/?utm_source=chatgpt.com\">Seeed JetPack 7.2 Deep Dive<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/flash_and_ota_jetpack_7.2\/?utm_source=chatgpt.com\">Seeed Flash &amp; OTA Guide<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetson_developtool_overview\/?utm_source=chatgpt.com\">Seeed Jetson DevelopTool<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/jetpack_7_2_memory_optimization_deep_dive\/?utm_source=chatgpt.com\">Seeed Memory Optimization Deep Dive<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/deploy_tensorrt_edge_llm_on_jetpack7.2\/?utm_source=chatgpt.com\">Seeed TensorRT Edge-LLM Guide<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/github.com\/Seeed-Studio\/wiki-documents\/blob\/docusaurus-version\/sites\/en\/docs\/Edge\/NVIDIA_Jetson\/JetPack_7_2\/AI_Inference\/JetPack_7_2_DeepStream.md?utm_source=chatgpt.com\">Seeed DeepStream 9.1 on JetPack 7.2<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/build_and_flash_yocto_for_seeed_jetson_carrier_boards\/?utm_source=chatgpt.com\">Seeed Yocto Guide<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/control_rebot_arm_with_nemoclaw_on_nvidia_jetson_thor\/?utm_source=chatgpt.com\">Seeed NemoClaw + reBot Arm Demo<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/industrial_vision_monitoring_on_industrial\/?utm_source=chatgpt.com\">Seeed Industrial Vision Demo<\/a><\/p>\n\n\n\n<p><a href=\"https:\/\/wiki.seeedstudio.com\/cn\/deploy_full_weight_gr00t_n1.7_tensorrt_jetpack7.2_agx_orin\/?utm_source=chatgpt.com\">Seeed GR00T N1.7 Guide<\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Conclusion<\/h2>\n\n\n\n<p>JetPack 7.2 is a significant step in the evolution of NVIDIA Jetson as an edge AI platform.<\/p>\n\n\n\n<p>Its importance does not come from one benchmark number or one new SDK.<\/p>\n\n\n\n<p>It comes from the combination of changes:<\/p>\n\n\n\n<p><strong>Ubuntu 24.04 + Linux 6.8 + <\/strong><strong>CUDA<\/strong><strong> 13.2 + updated TensorRT + DeepStream 9.1 + Agent Skills + NemoClaw + memory <\/strong><strong>optimization<\/strong><strong> + Yocto + new platform capabilities.<\/strong><\/p>\n\n\n\n<p>For developers, the immediate benefit is a newer foundation for AI workloads.<\/p>\n\n\n\n<p>For robotics teams, it creates a stronger path toward agentic and physical AI.<\/p>\n\n\n\n<p>For product teams, it provides more tools for turning a development board into a reproducible deployment platform.<\/p>\n\n\n\n<p>And for existing Jetson users, it creates a clear choice:<\/p>\n\n\n\n<p><strong>stay on a stable, validated software stack\u2014or invest in migration to unlock the capabilities of the next generation of edge AI.<\/strong><\/p>\n\n\n\n<p>The right way to make that decision is not to look at the release notes alone.<\/p>\n\n\n\n<p><strong>Build. Measure. Validate. Then deploy.<\/strong><\/p>\n\n\n\n<p><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udc65 Join Our Embodied AI Community<\/h2>\n\n\n\n<p>If you&#8217;re interested in:<\/p>\n\n\n\n<p><strong>Physical AI | Edge Computing | Agentic Robotics | Open-Source Robotic <\/strong><strong>Arm<\/strong><strong> | LeRobot | Robot Teleoperation | Robot Vision | VLA<\/strong><\/p>\n\n\n\n<p>Join our <strong>Embodied AI Community<\/strong> to connect, collaborate, and explore with developers worldwide.<\/p>\n\n\n\n<p><strong>Connect with us:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/discord.gg\/9PQnDAHmkK\">Discord<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/www.facebook.com\/groups\/rebotarm\">Facebook Group<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>A practical guide to JetPack 7.2, from the new software foundation and agentic AI workflows<\/p>\n","protected":false},"author":3703,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_lmt_disableupdate":"","_lmt_disable":"","_price":"","_stock":"","_tribe_ticket_header":"","_tribe_default_ticket_provider":"","_tribe_ticket_capacity":"0","_ticket_start_date":"","_ticket_end_date":"","_tribe_ticket_show_description":"","_tribe_ticket_show_not_going":false,"_tribe_ticket_use_global_stock":"","_tribe_ticket_global_stock_level":"","_global_stock_mode":"","_global_stock_cap":"","_tribe_rsvp_for_event":"","_tribe_ticket_going_count":"","_tribe_ticket_not_going_count":"","_tribe_tickets_list":"[]","_tribe_ticket_has_attendee_info_fields":false,"iawp_total_views":0,"footnotes":""},"categories":[1],"tags":[],"class_list":["post-132410","post","type-post","status-publish","format-standard","hentry","category-news"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v24.0 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson\" \/>\n<meta property=\"og:description\" content=\"A practical guide to JetPack 7.2, from the new software foundation and agentic AI workflows\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/\" \/>\n<meta property=\"og:site_name\" content=\"Latest News from Seeed Studio\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-04T11:24:26+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-09-04T12:05:34+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1672\" \/>\n\t<meta property=\"og:image:height\" content=\"941\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Frank\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Frank\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"16 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/\",\"url\":\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/\",\"name\":\"JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson\",\"isPartOf\":{\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-1030x580.png\",\"datePublished\":\"2026-09-04T11:24:26+00:00\",\"dateModified\":\"2026-09-04T12:05:34+00:00\",\"author\":{\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/#\/schema\/person\/1dd116f442e8ed65ddbd35b6d3f28d8a\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#primaryimage\",\"url\":\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1.png\",\"contentUrl\":\"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1.png\",\"width\":1672,\"height\":941},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/www.seeedstudio.com\/blog\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/#website\",\"url\":\"https:\/\/www.seeedstudio.com\/blog\/\",\"name\":\"Latest News from Seeed Studio\",\"description\":\"Emerging IoT, AI and Autonomous Applications on the Edge\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.seeedstudio.com\/blog\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/#\/schema\/person\/1dd116f442e8ed65ddbd35b6d3f28d8a\",\"name\":\"Frank\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.seeedstudio.com\/blog\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/secure.gravatar.com\/avatar\/fa87c178e5d477dfd0a9129d4a837e64?s=96&r=g\",\"contentUrl\":\"https:\/\/secure.gravatar.com\/avatar\/fa87c178e5d477dfd0a9129d4a837e64?s=96&r=g\",\"caption\":\"Frank\"},\"url\":\"https:\/\/www.seeedstudio.com\/blog\/author\/frank\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/","og_locale":"en_US","og_type":"article","og_title":"JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson","og_description":"A practical guide to JetPack 7.2, from the new software foundation and agentic AI workflows","og_url":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/","og_site_name":"Latest News from Seeed Studio","article_published_time":"2026-09-04T11:24:26+00:00","article_modified_time":"2026-09-04T12:05:34+00:00","og_image":[{"width":1672,"height":941,"url":"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1.png","type":"image\/png"}],"author":"Frank","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Frank","Est. reading time":"16 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/","url":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/","name":"JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson","isPartOf":{"@id":"https:\/\/www.seeedstudio.com\/blog\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#primaryimage"},"image":{"@id":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#primaryimage"},"thumbnailUrl":"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1-1030x580.png","datePublished":"2026-09-04T11:24:26+00:00","dateModified":"2026-09-04T12:05:34+00:00","author":{"@id":"https:\/\/www.seeedstudio.com\/blog\/#\/schema\/person\/1dd116f442e8ed65ddbd35b6d3f28d8a"},"breadcrumb":{"@id":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#primaryimage","url":"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1.png","contentUrl":"https:\/\/www.seeedstudio.com\/blog\/wp-content\/uploads\/2026\/09\/Jepack-7.2-Blog_1.png","width":1672,"height":941},{"@type":"BreadcrumbList","@id":"https:\/\/www.seeedstudio.com\/blog\/2026\/09\/04\/jetpack-7-2-guide\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.seeedstudio.com\/blog\/"},{"@type":"ListItem","position":2,"name":"JetPack 7.2: What&#8217;s New, Performance, Migration, and AI Deployment on NVIDIA Jetson"}]},{"@type":"WebSite","@id":"https:\/\/www.seeedstudio.com\/blog\/#website","url":"https:\/\/www.seeedstudio.com\/blog\/","name":"Latest News from Seeed Studio","description":"Emerging IoT, AI and Autonomous Applications on the Edge","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.seeedstudio.com\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/www.seeedstudio.com\/blog\/#\/schema\/person\/1dd116f442e8ed65ddbd35b6d3f28d8a","name":"Frank","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.seeedstudio.com\/blog\/#\/schema\/person\/image\/","url":"https:\/\/secure.gravatar.com\/avatar\/fa87c178e5d477dfd0a9129d4a837e64?s=96&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/fa87c178e5d477dfd0a9129d4a837e64?s=96&r=g","caption":"Frank"},"url":"https:\/\/www.seeedstudio.com\/blog\/author\/frank\/"}]}},"modified_by":"Frank","views":9,"featured_image_urls":{"full":"","thumbnail":"","medium":"","medium_large":"","large":"","1536x1536":"","2048x2048":"","visody_icon":"","magazine-7-slider-full":"","magazine-7-slider-center":"","magazine-7-featured":"","magazine-7-medium":"","magazine-7-medium-square":""},"author_info":{"display_name":"Frank","author_link":"https:\/\/www.seeedstudio.com\/blog\/author\/frank\/"},"category_info":"<a href=\"https:\/\/www.seeedstudio.com\/blog\/category\/news\/\" rel=\"category tag\">News<\/a>","tag_info":"News","comment_count":"0","_links":{"self":[{"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/posts\/132410","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/users\/3703"}],"replies":[{"embeddable":true,"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/comments?post=132410"}],"version-history":[{"count":4,"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/posts\/132410\/revisions"}],"predecessor-version":[{"id":132445,"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/posts\/132410\/revisions\/132445"}],"wp:attachment":[{"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/media?parent=132410"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/categories?post=132410"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.seeedstudio.com\/blog\/wp-json\/wp\/v2\/tags?post=132410"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}