{ "version": "2026-09-07", "description": "NVIDIA Jetson software configuration facts for the Jetson Configuration Checker: JetPack ↔ Jetson Linux (L4T) releases with CUDA / TensorRT / cuDNN / DeepStream / Ubuntu versions and supported modules, install methods and their limits, Super-mode enablement, on-device verify commands with expected output, runtime ↔ JetPack pairings and known issues with remediation. Every fact carries a provenance record (NVIDIA archive / release notes / developer guide = class A; developer-forum staff replies = class A; community examples = class C).", "last_updated": "2026-09-07", "confidence": "high", "evidence_class": "A", "module_ids": [ "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_thor_t5000", "jetson_thor_t4000" ], "legacy_module_ids": [ "jetson_xavier_nx", "jetson_agx_xavier", "jetson_nano" ], "schema": { "releases.": "{ jetpack, major, l4t, l4t_versions[], release_date, cuda, tensorrt, cudnn, vpi, deepstream, ubuntu, supported_modules[], notes, sources[] }", "install_methods.": "{ label, applies_to[], jetpack_range, notes, sources[] }", "verify_commands.": "{ command, expected_pattern, meaning, sources[] }", "runtime_compat[]": "{ runtime, jetpack, jetpack_min, version, install, notes, researched, sources[] }", "known_issues.": "{ modules[], jetpack, symptom, cause, remediation, sources[] }", "power_presets": "see datasets/power-modes.json (module × JetPack × nvpmodel preset)" }, "releases": { "5.1.2": { "jetpack": "5.1.2", "major": 5, "l4t": "35.4.1", "l4t_versions": [ "35.4.1" ], "release_date": "2023-08-25", "release_date_note": null, "cuda": "11.4.19", "tensorrt": "8.5.2", "cudnn": "8.6.0", "vpi": "2.3", "deepstream": null, "deepstream_note": null, "ubuntu": "20.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_xavier_nx", "jetson_agx_xavier" ], "notes": "First JetPack 5.1.x release to add Jetson AGX Orin Industrial support.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 5.1.2 Release Notes", "doc_id": "nvidia-docs:jetson/jetpack-archived/jetpack-512", "url": "https://docs.nvidia.com/jetson/archives/jetpack-archived/jetpack-512/release-notes/index.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "This release supports all Jetson AGX Orin, Jetson Orin NX, Jetson Orin Nano ... production modules." } ] }, "5.1.3": { "jetpack": "5.1.3", "major": 5, "l4t": "35.5.0", "l4t_versions": [ "35.5.0" ], "release_date": null, "release_date_note": null, "cuda": "11.4.19", "tensorrt": "8.5.2", "cudnn": "8.6.0", "vpi": "2.4", "deepstream": null, "deepstream_note": null, "ubuntu": "20.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_xavier_nx", "jetson_agx_xavier" ], "notes": "Minor update to 5.1.2; same compute stack (CUDA/TensorRT/cuDNN unchanged), VPI bumped to 2.4.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 5.1.3 Release Notes", "doc_id": "nvidia-docs:jetson/jetpack-archived/jetpack-513", "url": "https://docs.nvidia.com/jetson/archives/jetpack-archived/jetpack-513/release-notes/index.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "JetPack 5.1.3 packages the same compute stack as JetPack 5.1.2 but updates Vision Programming Interface (VPI) to VPI 2.4." } ] }, "5.1.4": { "jetpack": "5.1.4", "major": 5, "l4t": "35.6.0", "l4t_versions": [ "35.6.0" ], "release_date": "2024-09-04", "release_date_note": null, "cuda": "11.4.19", "tensorrt": "8.5.2", "cudnn": "8.6.0", "vpi": "2.4", "deepstream": null, "deepstream_note": null, "ubuntu": "20.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_xavier_nx", "jetson_agx_xavier" ], "notes": "Compute stack unchanged from 5.1.2/5.1.3.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 5.1.4 Release Notes", "doc_id": "nvidia-docs:jetson/jetpack-archived/jetpack-514", "url": "https://docs.nvidia.com/jetson/archives/jetpack-archived/jetpack-514/release-notes/index.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "L4T 35.6.0 ... Ubuntu 20.04 based root file system, ... Linux Kernel 5.10." } ] }, "5.1.5": { "jetpack": "5.1.5", "major": 5, "l4t": "35.6.1", "l4t_versions": [ "35.6.1" ], "release_date": "2025-03-12", "release_date_note": null, "cuda": "11.4.19", "tensorrt": "8.5.2", "cudnn": "8.6.0", "vpi": "2.4", "deepstream": null, "deepstream_note": null, "ubuntu": "20.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_xavier_nx", "jetson_agx_xavier" ], "notes": "JetPack archive also lists an L4T 35.6.2 revision under the 5.1.5 label — treat 35.6.1/35.6.2 as the same JetPack 5.1.5 point release with two BSP revisions (unresolved which is canonical; see gaps).", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 5.1.5 Release Notes", "doc_id": "nvidia-docs:jetson/jetpack-archived/jetpack-515", "url": "https://docs.nvidia.com/jetson/archives/jetpack-archived/jetpack-515/release-notes/index.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Jetson AGX Orin, Jetson Orin NX, Jetson Orin Nano, Jetson AGX Xavier and Jetson Xavier NX production modules." } ] }, "5.1.6": { "jetpack": "5.1.6", "major": 5, "l4t": "35.6.4", "l4t_versions": [ "35.6.4" ], "release_date": null, "release_date_note": null, "cuda": null, "tensorrt": null, "cudnn": null, "vpi": null, "deepstream": null, "deepstream_note": null, "ubuntu": "20.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_xavier_nx", "jetson_agx_xavier" ], "notes": "Listed on the JetPack Archive page; compute-stack versions (CUDA/TensorRT/cuDNN) not independently verified for this point release — gap.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack Archive", "doc_id": "nvidia-dev:embedded/jetpack-archive", "url": "https://developer.nvidia.com/embedded/jetpack-archive", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "JetPack 5.1.6 | L4T 35.6.4 — Supports: Jetson AGX Orin Series, Jetson Orin NX Series, Jetson Orin Nano Series, Jetson Xavier NX series, Jetson AGX" } ] }, "5.1.7": { "jetpack": "5.1.7", "major": 5, "l4t": "35.6.5", "l4t_versions": [ "35.6.5" ], "release_date": null, "release_date_note": null, "cuda": null, "tensorrt": null, "cudnn": null, "vpi": null, "deepstream": null, "deepstream_note": null, "ubuntu": "20.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_xavier_nx", "jetson_agx_xavier" ], "notes": "Latest 5.1.x point release per the JetPack Archive; compute-stack versions not independently verified — gap.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack Archive", "doc_id": "nvidia-dev:embedded/jetpack-archive", "url": "https://developer.nvidia.com/embedded/jetpack-archive", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "JetPack 5.1.7 | L4T 35.6.5 — Supports: Jetson AGX Orin Series, Jetson Orin NX Series, Jetson Orin Nano Series, Jetson Xavier NX series, Jetson AGX" } ] }, "6.0": { "jetpack": "6.0", "major": 6, "l4t": "36.3", "l4t_versions": [ "36.3" ], "release_date": "2024-05-30", "release_date_note": null, "cuda": "12.2.12", "tensorrt": "8.6.2", "cudnn": "8.9.4", "vpi": "3.1", "deepstream": null, "deepstream_note": null, "ubuntu": "22.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb" ], "notes": "First JetPack 6 production release. Requires reflash from JetPack 5 (see known_issues jp5_to_jp6_no_ota).", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 6.0 Release Notes", "doc_id": "nvidia-docs:jetson/jetpack-archived/jetpack-60", "url": "https://docs.nvidia.com/jetson/archives/jetpack-archived/jetpack-60/release-notes/index.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Jetson Linux 36.3 which packs Linux Kernel 5.15 and Ubuntu 22.04 based root file system." } ] }, "6.1": { "jetpack": "6.1", "major": 6, "l4t": "36.4", "l4t_versions": [ "36.4" ], "release_date": "2024-12-16", "release_date_note": null, "cuda": "12.6.10", "tensorrt": "10.3.0", "cudnn": "9.3.0", "vpi": "3.2", "deepstream": "7.1", "deepstream_note": null, "ubuntu": "22.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb" ], "notes": "Adds Jetson Orin Nano Developer Kit MAXN boost mode (predecessor to the JetPack 6.2 Super flash config).", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 6.1 Release Notes", "doc_id": "nvidia-docs:jetson/jetpack-archived/jetpack-61", "url": "https://docs.nvidia.com/jetson/archives/jetpack-archived/jetpack-61/release-notes/index.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "JetPack 6.1 supports all NVIDIA Jetson Orin modules and developer kits." } ] }, "6.2": { "jetpack": "6.2", "major": 6, "l4t": "36.4.3", "l4t_versions": [ "36.4.3" ], "release_date": "2025-01-15", "release_date_note": null, "cuda": "12.6.10", "tensorrt": "10.3.0", "cudnn": "9.3.0", "vpi": "3.2", "deepstream": null, "deepstream_note": null, "ubuntu": "22.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb" ], "notes": "Introduces Super Mode (MAXN_SUPER + higher wattage presets) for Orin Nano and Orin NX, gated by the jetson-orin-nano-devkit-super flash configuration.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 6.2 Release Notes", "doc_id": "nvidia-docs:jetson/jetpack-archived/jetpack-62", "url": "https://docs.nvidia.com/jetson/archives/jetpack-archived/jetpack-62/release-notes/index.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "NVIDIA Jetson Linux 36.4.3 supports all NVIDIA Jetson Orin modules and developer kits." } ] }, "6.2.1": { "jetpack": "6.2.1", "major": 6, "l4t": "36.4.4", "l4t_versions": [ "36.4.4" ], "release_date": "2025-06-26", "release_date_note": null, "cuda": "12.6.10", "tensorrt": "10.3.0", "cudnn": "9.3.0", "vpi": "3.2", "deepstream": null, "deepstream_note": null, "ubuntu": "22.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb" ], "notes": "Minor BSP update; adds Hardware Security Module (HSM) boot-image signing support and Docker v28.0.x compatibility. Same compute stack as 6.2.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 6.2.1 Release Notes (current)", "doc_id": "nvidia-docs:jetson/jetpack/release-notes", "url": "https://docs.nvidia.com/jetson/jetpack/release-notes/index.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "NVIDIA Jetson Linux 36.4.4 is a minor improvement over Jetson Linux 36.4.3." } ] }, "6.2.2": { "jetpack": "6.2.2", "major": 6, "l4t": "36.5.0", "l4t_versions": [ "36.5.0" ], "release_date": null, "release_date_note": null, "cuda": null, "tensorrt": null, "cudnn": null, "vpi": null, "deepstream": null, "deepstream_note": null, "ubuntu": "22.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb" ], "notes": "Listed on JetPack Archive; component versions (CUDA/TensorRT/cuDNN) not independently verified for this point release — gap. L4T mapping matches edgeaistack-api/src/v2/data/power_modes_data.js JETPACK_L4T.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack Archive", "doc_id": "nvidia-dev:embedded/jetpack-archive", "url": "https://developer.nvidia.com/embedded/jetpack-archive", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "JetPack 6.2.2 | L4T 36.5.0 — Supports: Jetson AGX Orin Series, Jetson Orin NX Series, Jetson Orin Nano Series" } ] }, "6.2.3": { "jetpack": "6.2.3", "major": 6, "l4t": "36.5.2", "l4t_versions": [ "36.5.2" ], "release_date": null, "release_date_note": null, "cuda": null, "tensorrt": null, "cudnn": null, "vpi": null, "deepstream": null, "deepstream_note": null, "ubuntu": "22.04", "supported_modules": [ "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_orin_nano", "jetson_orin_nano_4gb" ], "notes": "Latest 6.2.x point release per the JetPack Archive; component versions not independently verified — gap.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack Archive", "doc_id": "nvidia-dev:embedded/jetpack-archive", "url": "https://developer.nvidia.com/embedded/jetpack-archive", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "JetPack 6.2.3 | L4T 36.5.2 — Supports: Jetson AGX Orin Series, Jetson Orin NX Series, Jetson Orin Nano Series" } ] }, "7.0": { "jetpack": "7.0", "major": 7, "l4t": "38.2", "l4t_versions": [ "38.2", "38.2.1" ], "release_date": "2025-08-25", "release_date_note": "2025-08-25 (38.2); 2025-09-18 (38.2.1)", "cuda": "13.0.0", "tensorrt": "10.13.2.6", "cudnn": "9.12.0", "vpi": "4.0.0", "deepstream": null, "deepstream_note": null, "ubuntu": "24.04", "supported_modules": [ "jetson_thor_t5000" ], "notes": "First JetPack 7 release; Thor-only (Orin family not yet supported by the 7.x line at this point). Aligns Jetson software with Server Base System Architecture (SBSA). Does NOT support T4000 (added in 7.1).", "sources": [ { "publisher": "NVIDIA", "document": "JetPack SDK Downloads and Notes Archive — 7.0", "doc_id": "nvidia-dev:embedded/jetpack/downloads/archive-7.0", "url": "https://developer.nvidia.com/embedded/jetpack/downloads/archive-7.0", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Jetson Linux 38.2 (38.2.1 is the latest 38.2 version released) ... CUDA 13.0.0 ... Jetson AGX Thor Developer Kit, Jetson T5000." } ] }, "7.1": { "jetpack": "7.1", "major": 7, "l4t": "38.4", "l4t_versions": [ "38.4" ], "release_date": "2026-01-06", "release_date_note": null, "cuda": "13.0.0", "tensorrt": "10.13.3.9", "cudnn": "9.12.0", "vpi": "4.0.0", "deepstream": "7.1", "deepstream_note": "7.1 (per DeepStream 7.1 release notes covering Jetson; not independently confirmed against 7.1 JetPack line — see gap)", "ubuntu": "24.04", "supported_modules": [ "jetson_thor_t5000", "jetson_thor_t4000" ], "notes": "First JetPack release to support Jetson T4000. Introduces TensorRT Edge-LLM (open-source C++ SDK for LLM/VLM inference on edge) and a unified Video Codec SDK.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack SDK Downloads and Notes Archive — 7.1 (Jetson Linux 38.4)", "doc_id": "nvidia-dev:embedded/jetpack/downloads/archive-7.1", "url": "https://developer.nvidia.com/embedded/jetpack/downloads/archive-7.1", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Jetson Linux 38.4 ... CUDA 13.0.0 ... Jetson AGX Thor Developer Kit, Jetson T5000, Jetson T4000." } ] }, "7.2": { "jetpack": "7.2", "major": 7, "l4t": "39.2", "l4t_versions": [ "39.2" ], "release_date": "2026-06-02", "release_date_note": null, "cuda": "13.2.1", "tensorrt": "10.16.2", "cudnn": "9.20.0", "vpi": "4.1.3", "deepstream": null, "deepstream_note": null, "ubuntu": "24.04", "supported_modules": [ "jetson_thor_t5000", "jetson_thor_t4000", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial" ], "notes": "First JetPack 7.x release to unify Orin family support alongside Thor. Removes the SD Card image for Jetson Orin Nano Developer Kit in favor of a single unified ISO image (see known_issues no_sd_card_jp72 and iso_thread in power_modes_data.js DOC).", "sources": [ { "publisher": "NVIDIA", "document": "JetPack SDK Downloads and Notes Archive — 7.2", "doc_id": "nvidia-dev:embedded/jetpack/downloads/archive-7.2", "url": "https://developer.nvidia.com/embedded/jetpack/downloads/archive-7.2", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "JetPack 7.2 with Jetson Linux 39.2 ... Jetson AGX Thor Developer Kit, Jetson T5000, Jetson T4000, Jetson Orin Family. There is no more SD Card" } ] }, "7.2.1": { "jetpack": "7.2.1", "major": 7, "l4t": "39.2.1", "l4t_versions": [ "39.2.1" ], "release_date": "2026-08-11", "release_date_note": null, "cuda": "13.2.1", "tensorrt": "10.16.2", "cudnn": "9.20.0", "vpi": "4.1.3", "deepstream": null, "deepstream_note": null, "ubuntu": "24.04", "supported_modules": [ "jetson_thor_t5000", "jetson_thor_t4000", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial" ], "notes": "Latest JetPack release as of retrieval. Adds 'agentic video skills' and Jetson T3000 emulation support (T3000 is NOT yet a shipping module — emulation only; not added to module_ids). Requires JetPack 6.x-generation Jetson UEFI/QSPI firmware on Orin Nano devkits before flashing (see known_issues jp72_orin_nano_firmware_prereq).", "sources": [ { "publisher": "NVIDIA", "document": "NVIDIA JetPack 7.2.1 Adds Agentic Video Skills and T3000 Emulation", "doc_id": "nvidia-dev:blog/nvidia-jetpack-7-2-1-adds-agentic-video-skills-and-t3000-emulation", "url": "https://developer.nvidia.com/blog/nvidia-jetpack-7-2-1-adds-agentic-video-skills-and-t3000-emulation/", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Jetson Linux 39.2.1 ... Support for Jetson T3000 emulation." } ] } }, "install_methods": { "sdk_manager": { "id": "sdk_manager", "label": "SDK Manager (host flash)", "applies_to": [ "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_thor_t5000", "jetson_thor_t4000", "jetson_xavier_nx", "jetson_agx_xavier" ], "jetpack_range": "all", "notes": "Host-driven flash from an x86 Ubuntu machine (or a Docker container per NVIDIA's SDK Manager Docker image). Only method that can flash the Super flash configuration and the JetPack 6.x/7.x QSPI-firmware bridge update; required to change flash configuration (e.g. move a non-Super Orin Nano to jetson-orin-nano-devkit-super).", "sources": [ { "publisher": "NVIDIA", "document": "SUPER_ENABLEMENT (edgeaistack-api power_modes_data.js, cross-referencing orin_guide)", "doc_id": "nvidia-docs:jetson/r36.4.4/DeveloperGuide", "url": "https://docs.nvidia.com/jetson/archives/r36.4.4/DeveloperGuide/SD/PlatformPowerAndPerformance/JetsonOrinNanoSeriesJetsonOrinNxSeriesAndJetsonAgxOrinSeries.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "MAXN_SUPER mode is available only when flashing with jetson-orin-nano-devkit-super or jetson-orin-nano-devkit-super-maxn configuration." } ] }, "sd_card_image": { "id": "sd_card_image", "label": "SD card image", "applies_to": [ "jetson_orin_nano" ], "jetpack_range": "5.1–7.1 (removed in 7.2)", "notes": "Only the Jetson Orin Nano Developer Kit carrier board has a microSD slot on the module; Orin NX and production Orin Nano modules have no SD/eMMC and require NVMe or SDK-Manager flashing instead. JetPack 7.2 discontinued the SD Card image entirely in favor of the unified ISO image.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack 7.2 Downloads and Notes", "doc_id": "nvidia-dev:embedded/jetpack/downloads/archive-7.2", "url": "https://developer.nvidia.com/embedded/jetpack/downloads/archive-7.2", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "There is no more SD Card image for Jetson Orin Nano Developer Kit. Use the unified ISO image to flash the developer kit using a" } ] }, "iso_image": { "id": "iso_image", "label": "ISO image (target-side)", "applies_to": [ "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_thor_t5000", "jetson_thor_t4000" ], "jetpack_range": "6.0+", "notes": "USB-stick installable image, flashed target-side (no x86 host required). Cannot upgrade a non-Super device to Super Mode (must reflash from an x86 host / SDK Manager, or hand-edit TNSPEC/COMPATIBLE_SPEC in /etc/nv_boot_control.conf). JetPack 7.2's ISO image is the sole install path for the Orin Nano Developer Kit now that the SD Card image is gone.", "sources": [ { "publisher": "NVIDIA", "document": "NVIDIA Developer Forums — 25W and MAXN SUPER not seen in JetPack 7.2", "doc_id": "forums.developer.nvidia.com/t/372627", "url": "https://forums.developer.nvidia.com/t/25w-and-maxn-super-not-seen-in-jetpack-7-2/372627", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Currently, the ISO image cannot upgrade a device from non-Super Mode to Super Mode." } ] }, "apt_upgrade": { "id": "apt_upgrade", "label": "apt upgrade / OTA", "applies_to": [ "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_thor_t5000", "jetson_thor_t4000" ], "jetpack_range": "minor/point releases within the same major L4T line only (e.g. 6.0→6.2.x, 7.0→7.2.x)", "notes": "In-place `apt update && apt dist-upgrade` (or NVIDIA's OTA image mechanism) works only within the same L4T major version. Crossing a major boundary (5.x→6.x, 6.x→7.x) is unsupported and requires a full reflash because the bootloader/kernel/rootfs and base Ubuntu version all change.", "sources": [ { "publisher": "NVIDIA", "document": "NVIDIA Developer Forums — Expected JetPack 5 to JetPack 6 Upgrade paths (staff reply, linuxdev)", "doc_id": "forums.developer.nvidia.com/t/280782", "url": "https://forums.developer.nvidia.com/t/expected-jetpack-5-to-jetpack-6-upgrade-paths/280782", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Correct, you won't be able to OTA update from 5 to 6. You'll be able to upgrade minor upgrades only." } ] }, "flash_sh": { "id": "flash_sh", "label": "flash.sh / Jetson Linux BSP", "applies_to": [ "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_thor_t5000", "jetson_thor_t4000", "jetson_xavier_nx", "jetson_agx_xavier" ], "jetpack_range": "all", "notes": "Command-line flashing (`sudo ./flash.sh mmcblk0p1` or `tools/kernel_flash` scripts) using the Jetson Linux BSP tarball + sample rootfs extracted on an x86 Ubuntu host, in recovery mode over USB. Underlies SDK Manager's GUI flash and is the only way to pass a custom board/flash configuration string (e.g. jetson-orin-nano-devkit-super) without the GUI.", "sources": [ { "publisher": "NVIDIA", "document": "Jetson Orin Nano Developer Kit User Guide — JetPack 6.x Update Path", "doc_id": "nvidia-docs:jetson/orin-nano-devkit/user-guide/latest/update_firmware.html", "url": "https://docs.nvidia.com/jetson/orin-nano-devkit/user-guide/latest/update_firmware.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Some developer kits shipped with factory firmware that cannot boot JetPack 6.x or JetPack 7.2.1." } ] } }, "super_mode": { "requirements": { "jetpack_min": "6.2", "flash_config": "jetson-orin-nano-devkit-super (or jetson-orin-nano-devkit-super-maxn)", "modules": [ "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb" ], "iso_limitation": "The ISO image cannot upgrade a device already flashed non-Super into Super Mode; must reflash from an x86 host via SDK Manager, or repair TNSPEC/COMPATIBLE_SPEC in /etc/nv_boot_control.conf to carry the -super suffix." }, "how_to_verify": [ { "command": "nvpmodel -q", "meaning": "Shows current NV Power Mode name and numeric mode ID; a Super-flashed Orin Nano/NX exposes MAXN_SUPER and a 25W/40W preset that a non-Super flash does not.", "expected_pattern": null }, { "command": "sudo nvpmodel -m ", "meaning": "Switches to a different nvpmodel preset by numeric mode ID (from the same table `nvpmodel -q` reads).", "expected_pattern": null }, { "command": "cat /proc/device-tree/model", "meaning": "A Super-flashed Orin Nano devkit reports a distinct device-tree model string containing 'Super'.", "expected_pattern": null } ], "sources": [ { "publisher": "NVIDIA", "document": "Jetson Linux Developer Guide — Platform Power and Performance (r36.4.4)", "doc_id": "nvidia-docs:jetson/r36.4.4/DeveloperGuide", "url": "https://docs.nvidia.com/jetson/archives/r36.4.4/DeveloperGuide/SD/PlatformPowerAndPerformance/JetsonOrinNanoSeriesJetsonOrinNxSeriesAndJetsonAgxOrinSeries.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "MAXN_SUPER mode is available only when flashing with jetson-orin-nano-devkit-super or jetson-orin-nano-devkit-super-maxn configuration." }, { "publisher": "NVIDIA", "document": "NVIDIA Developer Forums — JetPack 6.2 Brings Super Mode", "doc_id": "forums.developer.nvidia.com/t/320344", "url": "https://forums.developer.nvidia.com/t/jetpack-6-2-brings-super-mode-to-nvidia-jetson-orin-nano-and-jetson-orin-nx-modules/320344", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Jetpack 6.2 brings Super Mode to NVIDIA Jetson Orin Nano and Jetson Orin NX modules." }, { "publisher": "NVIDIA", "document": "NVIDIA Developer Forums — 25W and MAXN SUPER not seen in JetPack 7.2", "doc_id": "forums.developer.nvidia.com/t/372627", "url": "https://forums.developer.nvidia.com/t/25w-and-maxn-super-not-seen-in-jetpack-7-2/372627", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Currently, the ISO image cannot upgrade a device from non-Super Mode to Super Mode." }, { "publisher": "community (medium.com)", "document": "cat /proc/device-tree/model example output on Orin Nano Super", "doc_id": "medium.com/@arnav04verma/bootstrapping-a-headless-jetson-orin-nano-super-for-uav-autonomy-", "url": "https://medium.com/@arnav04verma/bootstrapping-a-headless-jetson-orin-nano-super-for-uav-autonomy-026b7c972dda", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "NVIDIA Jetson Orin Nano Engineering Reference Developer Kit Super" } ] }, "verify_commands": { "l4t_release": { "id": "l4t_release", "command": "cat /etc/nv_tegra_release", "expected_pattern": "# R36 (release), REVISION: 4.3, GCID: , BOARD: generic, EABI: aarch64, DATE: ", "meaning": "Reports the L4T major release (R##) and REVISION (##.#), which maps 1:1 to a JetPack version via the JETPACK_L4T table. NVIDIA has flagged this file as legacy on newer L4T releases in favor of dpkg-based checks.", "sources": [ { "publisher": "community (aggregated forum examples)", "document": "NVIDIA Developer Forums — Jetson JetPack version verification threads", "doc_id": "forums.developer.nvidia.com/t/379754", "url": "https://forums.developer.nvidia.com/t/verifying-whether-the-jetpack-version-is-jetpack6-2/379754", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "R36 (release), REVISION: 4.3, GCID: 38968081, BOARD: generic, EABI: aarch64, DATE: Wed Jan 8 01:49:37 UTC 2025" } ] }, "jetpack_apt": { "id": "jetpack_apt", "command": "dpkg -l nvidia-jetpack (or: apt-cache show nvidia-jetpack; dpkg-query --show nvidia-l4t-core)", "expected_pattern": "ii nvidia-jetpack arm64 NVIDIA Jetpack Meta Package", "meaning": "Reports the installed nvidia-jetpack meta-package version — only present if the full JetPack SDK (not just the BSP) was installed via SDK Manager or apt.", "sources": [ { "publisher": "community (Collabnix / NVIDIA forum aggregation)", "document": "How to find Jetpack version of NVIDIA Jetson Nano", "doc_id": "forums.developer.nvidia.com/t/145422", "url": "https://forums.developer.nvidia.com/t/how-do-you-check-confirm-jetpack-version-on-the-command-line/145422", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "dpkg -l nvidia-jetpack ... if the full SDK was installed via SDK Manager." } ] }, "jetson_release": { "id": "jetson_release", "command": "jetson_release -v (from the jetson-stats / jtop package)", "expected_pattern": "Model: - Jetpack [L4T ]", "meaning": "Third-party (rbonghi/jetson_stats) tool that decodes hardware model, L4T, and JetPack version together, plus power mode and library versions; requires `sudo pip3 install -U jetson-stats` and a reboot.", "sources": [ { "publisher": "community (jetson_stats project via forum/GitHub)", "document": "jetson_release -v example output", "doc_id": "github:rbonghi/jetson_stats", "url": "https://github.com/rbonghi/jetson_stats", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "Model: Jetson AGX Orin Developer Kit - Jetpack 5.1.2 [L4T 35.4.1]" } ] }, "nvcc_version": { "id": "nvcc_version", "command": "nvcc --version", "expected_pattern": "Cuda compilation tools, release , V", "meaning": "Reports the installed CUDA toolkit version; requires CUDA toolkit (not just the driver/runtime) to be installed and /usr/local/cuda/bin on PATH.", "sources": [ { "publisher": "community (jetsonhacks/jetsonUtilities aggregation)", "document": "different ways to check for nvidia jetpack versions", "doc_id": "github:jadwigo/86b905ca2573dc7b9a685652b82ef590", "url": "https://gist.github.com/jadwigo/86b905ca2573dc7b9a685652b82ef590", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "nvcc --version" } ] }, "tensorrt_dpkg": { "id": "tensorrt_dpkg", "command": "dpkg -l | grep tensorrt", "expected_pattern": "ii tensorrt arm64 ... (and libnvinfer*, python3-libnvinfer* rows)", "meaning": "Lists the installed TensorRT meta-package and libnvinfer runtime/dev packages with exact version strings.", "sources": [ { "publisher": "community (aggregated Jetson verification guides)", "document": "different ways to check for nvidia jetpack versions", "doc_id": "github:jadwigo/86b905ca2573dc7b9a685652b82ef590", "url": "https://gist.github.com/jadwigo/86b905ca2573dc7b9a685652b82ef590", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "dpkg -l | grep -i 'jetpack'" } ] }, "tensorrt_python": { "id": "tensorrt_python", "command": "python3 -c \"import tensorrt; print(tensorrt.__version__)\"", "expected_pattern": " (e.g. 10.3.0)", "meaning": "Confirms the Python TensorRT bindings are importable and reports their version; fails with ModuleNotFoundError if only the C++ TensorRT libraries were installed without the python3-libnvinfer bindings.", "sources": [ { "publisher": "community", "document": "General Jetson TensorRT verification pattern (standard Python idiom, cross-checked against NVIDIA TensorRT docs)", "doc_id": "nvidia-docs:deeplearning/tensorrt/latest/getting-started/release-notes-10/10.13.2.html", "url": "https://docs.nvidia.com/deeplearning/tensorrt/latest/getting-started/release-notes-10/10.13.2.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "python3 -c \"import tensorrt\"" } ] }, "nvpmodel_query": { "id": "nvpmodel_query", "command": "sudo nvpmodel -q", "expected_pattern": "NV Power Mode: \n", "meaning": "Reports the active nvpmodel preset name and numeric ID. `nvpmodel -q --verbose` additionally prints per-mode CPU online count and CPU/GPU clock ceilings.", "sources": [ { "publisher": "community (JetsonHacks / RidgeRun aggregation)", "document": "NVPModel usage examples", "doc_id": "jetsonhacks.com/2018/10/07/nvpmodel-nvidia-jetson-agx-xavier-developer-kit", "url": "https://jetsonhacks.com/2018/10/07/nvpmodel-nvidia-jetson-agx-xavier-developer-kit/", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "NV Power Mode: MAXN 0" } ] }, "jetson_clocks_show": { "id": "jetson_clocks_show", "command": "sudo jetson_clocks --show", "expected_pattern": "SOC family: Machine:\nOnline CPUs: \ncpu: Online=<0|1> Governor= MinFreq= MaxFreq= CurrentFreq=\nGPU MinFreq= ... MaxFreq= ... CurrentFreq=...", "meaning": "Shows current vs. max CPU/GPU/EMC clocks; running `jetson_clocks` (no --show) locks clocks to the current nvpmodel preset's maximum, which is how NVIDIA's own benchmark numbers are captured.", "sources": [ { "publisher": "NVIDIA", "document": "Jetson Linux Developer Guide (general jetson_clocks usage, cross-referenced from power_modes_data.js sources)", "doc_id": "nvidia-docs:jetson/r36.4.4/DeveloperGuide", "url": "https://docs.nvidia.com/jetson/archives/r36.4.4/DeveloperGuide/SD/PlatformPowerAndPerformance/JetsonOrinNanoSeriesJetsonOrinNxSeriesAndJetsonAgxOrinSeries.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "B", "quote": "jetson_clocks --show" } ] }, "tegrastats": { "id": "tegrastats", "command": "tegrastats", "expected_pattern": "RAM /MB ... CPU [%@,...] ... GR3D_FREQ %@ ... VDD_* /", "meaning": "Live streaming telemetry (RAM, per-core CPU %/freq, GPU %/freq, per-rail power in mW) sampled every N ms; used to confirm a power budget/thermal setting is actually taking effect.", "sources": [ { "publisher": "NVIDIA", "document": "NVIDIA Developer Forums — JetPack 6: tegrastats missing GPU temperature", "doc_id": "forums.developer.nvidia.com/t/309539", "url": "https://forums.developer.nvidia.com/t/jetpack-6-tegrastats-missing-gpu-temperature/309539", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "tegrastats" } ] }, "device_tree_model": { "id": "device_tree_model", "command": "cat /proc/device-tree/model", "expected_pattern": "NVIDIA Jetson Developer Kit (Super-flashed boards append 'Super' or 'Engineering Reference Developer Kit Super')", "meaning": "Identifies the exact board/carrier and, on Orin Nano, distinguishes a Super flash configuration from a standard one.", "sources": [ { "publisher": "community", "document": "Bootstrapping a Headless Jetson Orin Nano Super for UAV Autonomy", "doc_id": "medium.com/@arnav04verma/bootstrapping-a-headless-jetson-orin-nano-super-for-uav-autonomy-", "url": "https://medium.com/@arnav04verma/bootstrapping-a-headless-jetson-orin-nano-super-for-uav-autonomy-026b7c972dda", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "NVIDIA Jetson Orin Nano Engineering Reference Developer Kit Super" } ] }, "nvhost_devices": { "id": "nvhost_devices", "command": "ls /dev/nvhost*", "expected_pattern": "/dev/nvhost-ctrl /dev/nvhost-ctrl-gpu /dev/nvhost-gpu /dev/nvhost-nvdec /dev/nvhost-nvenc /dev/nvhost-vic ...", "meaning": "Confirms the NVIDIA host1x/nvhost kernel driver stack (GPU, NVDEC, NVENC, VIC) is loaded and device nodes exist; a missing nvhost-nvdec/nvenc node indicates the hardware codec driver failed to load.", "sources": [] }, "deepstream_version": { "id": "deepstream_version", "command": "deepstream-app --version-all", "expected_pattern": "deepstream-app version \nDeepStreamSDK \nCUDA Driver Version: \nCUDA Runtime Version: \nTensorRT Version: \ncuDNN Version: ", "meaning": "Reports DeepStream SDK version plus the CUDA/TensorRT/cuDNN versions DeepStream itself was built against, for cross-checking against `nvcc --version` / `dpkg -l | grep tensorrt`.", "sources": [ { "publisher": "community", "document": "deepstream-app --version-all example output (DeepStream 7.1.0)", "doc_id": "nvidia-docs:metropolis/deepstream/dev-guide/text/DS_ref_app_deepstream.html", "url": "https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_ref_app_deepstream.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "deepstream-app version 7.1.0 DeepStreamSDK 7.1.0 CUDA Driver Version: 12.6 CUDA Runtime Version: 12.6 TensorRT Version: 10.3 cuDNN Version: 9.0" } ] }, "docker_nvidia_runtime": { "id": "docker_nvidia_runtime", "command": "sudo dpkg --get-selections | grep nvidia ; sudo docker info | grep -i runtime", "expected_pattern": "nvidia-container-runtime install\nnvidia-container-runtime-hook install\nnvidia-docker2 install\n...\nRuntimes: nvidia runc", "meaning": "Confirms the NVIDIA Container Toolkit packages are installed and Docker has registered the `nvidia` runtime (required for `--runtime nvidia` / GPU-enabled containers, including jetson-containers and DeepStream/L4T container images).", "sources": [ { "publisher": "NVIDIA (community-maintained wiki mirror of NVIDIA container docs)", "document": "NVIDIA Container Runtime on Jetson", "doc_id": "github:Technology-Notes/nvidia-docker-wiki", "url": "https://github.com/Technology-Notes/nvidia-docker-wiki/blob/master/NVIDIA-Container-Runtime-on-Jetson.md", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "sudo dpkg --get-selections | grep nvidia ... sudo docker info | grep nvidia ... + Runtimes: nvidia runc" } ] }, "ollama_llamacpp_cuda": { "id": "ollama_llamacpp_cuda", "command": "ollama run --verbose (look for 'looking for compatible GPU'); or: ldd $(which llama-cli) | grep cuda", "expected_pattern": "no single canonical NVIDIA-documented pattern found", "meaning": "Both ollama and llama.cpp print a GPU-offload confirmation line at startup on a CUDA build, but the exact wording is version-dependent community behavior, not an NVIDIA-documented contract.", "sources": [] } }, "runtime_compat": [ { "runtime": "pytorch", "jetpack": "6.2", "jetpack_min": "6.2", "version": "2.8.0 / 0.23.0 (torchvision)", "install": "python -m pip install torch==2.8.0 torchvision==0.23.0 --index-url=https://pypi.jetson-ai-lab.dev/jp6/cu126", "notes": "NVIDIA does not publish PyTorch wheels through pip directly for Jetson (aarch64); the Jetson AI Lab community index (pypi.jetson-ai-lab.dev, mirrored at .io) hosts JetPack-version-pinned wheel sets (e.g. /jp6/cu126 for JetPack 6.2). PyTorch.org wheels are x86_64-only and will not install on Jetson.", "researched": true, "sources": [ { "publisher": "community (Jetson AI Lab / forum aggregation)", "document": "Compatible pytorch version for the jetson orin with jetpack 6.2", "doc_id": "forums.developer.nvidia.com/t/325045", "url": "https://forums.developer.nvidia.com/t/compatible-pytorch-version-for-the-jetson-orin-with-jetpack-6-2/325045/10", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "python -m pip install torch==2.8.0 torchvision==0.23.0 --index-url=https://pypi.jetson-ai-lab.io/jp6/cu126" } ] }, { "runtime": "jetson_containers", "jetpack": "5.1+ / 6.x", "jetpack_min": "5.1", "version": "tag scheme: -r, e.g. 2.1-r36.2.0", "install": "docker run via the jetson-containers `run`/`autotag` scripts, or `docker pull dustynv/l4t-pytorch:r36.2.0`", "notes": "dusty-nv/jetson-containers builds and tags container images per L4T minor line; images built for one L4T R35.x point release are compatible with other R35.x releases (same for R36.x), and `autotag` auto-selects a tag matching the host's detected JetPack/L4T.", "researched": true, "sources": [ { "publisher": "community (dusty-nv/jetson-containers project)", "document": "jetson-containers README / package docs", "doc_id": "github:dusty-nv/jetson-containers", "url": "https://github.com/dusty-nv/jetson-containers", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "L4T R32.7 containers can run on other versions of L4T R32.7 (JetPack 4.6+) and L4T R35.x containers can run on other versions of L4T R35.x" } ] }, { "runtime": "deepstream", "jetpack": "6.1", "jetpack_min": "6.1", "version": "7.1", "install": "apt install deepstream-7.1 (via NVIDIA DeepStream apt repo) or the deepstream-l4t container image matched to L4T 36.4", "notes": "NVIDIA staff confirmed DeepStream 7.1's officially supported/compatible JetPack version is 6.1, not 6.2; forum users report functional but unofficial use on 6.2 with no compatibility guarantee for future JetPack releases.", "researched": true, "sources": [ { "publisher": "NVIDIA", "document": "NVIDIA Developer Forums — Deepstream 7.1 compatibility with Jetpack 6.2 and MAXN mode (staff reply, yuweiw)", "doc_id": "forums.developer.nvidia.com/t/325451", "url": "https://forums.developer.nvidia.com/t/deepstream-7-1-compatibility-with-jetpack-6-2-and-maxn-mode/325451", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "The compatible Jetpack Version for DeepStream 7.1 is Jetpack 6.1. We will update the DeepStream version as planned in the future, but there is no" } ] }, { "runtime": "tensorrt_llm", "jetpack": "7.1", "jetpack_min": "7.1", "version": "TensorRT Edge-LLM (not mainline TensorRT-LLM)", "install": "Jetson AI Lab tutorial-driven build (CUDA 13.2 toolkit + TensorRT dev packages); see jetson-ai-lab.com/tutorials/tensorrt-edge-llm", "notes": "NVIDIA ships a Jetson-specific 'TensorRT Edge-LLM' C++ SDK on Thor starting JetPack 7.1 rather than the datacenter TensorRT-LLM package; positioned for on-device LLM/VLM inference on T4000/T5000, not for Orin.", "researched": true, "sources": [ { "publisher": "NVIDIA", "document": "Accelerate AI Inference for Edge and Robotics with NVIDIA Jetson T4000 and NVIDIA JetPack 7.1", "doc_id": "nvidia-dev:blog/accelerate-ai-inference-for-edge-and-robotics-with-nvidia-jetson-t4000-and-", "url": "https://developer.nvidia.com/blog/accelerate-ai-inference-for-edge-and-robotics-with-nvidia-jetson-t4000-and-nvidia-jetpack-7-1/", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "B", "quote": "JetPack 7.1 introduces TensorRT Edge-LLM, an open-source C++ SDK that enables efficient LLM and VLM inference on edge platforms." } ] }, { "runtime": "vllm", "jetpack": "7.0+ (Thor only)", "jetpack_min": "7.0", "version": "latest (rolling container tag)", "install": "docker pull ghcr.io/nvidia-ai-iot/vllm:latest-jetson-thor (Orin: ghcr.io/nvidia-ai-iot/vllm:latest-jetson-orin)", "notes": "vLLM is distributed as prebuilt containers per Jetson family (Thor vs. Orin), not as a pip wheel; the Thor tag targets JetPack 7.x, the Orin tag targets JetPack 6.x.", "researched": true, "sources": [ { "publisher": "community (Jetson AI Lab / NVIDIA-AI-IOT GitHub org, referenced via NVIDIA blog ecosystem coverage)", "document": "Jetson AI Lab vLLM container references", "doc_id": "www.jetson-ai-lab.com/tutorials/genai-on-jetson-llms-vlms", "url": "https://www.jetson-ai-lab.com/tutorials/genai-on-jetson-llms-vlms/", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "Prebuilt vLLM containers are available for Jetson Thor (ghcr.io/nvidia-ai-iot/vllm:latest-jetson-thor) and Jetson Orin (ghcr.io/nvidia-ai-iot/vllm:latest-jetson-orin)." } ] }, { "runtime": "onnxruntime", "jetpack": null, "jetpack_min": null, "version": null, "install": "", "notes": "Not independently researched this pass — gap. Historically distributed via Jetson Zoo / jetson-containers `onnxruntime` package with a TensorRT execution provider build per JetPack line.", "researched": false, "sources": [] }, { "runtime": "llama_cpp", "jetpack": null, "jetpack_min": null, "version": null, "install": "", "notes": "Not independently researched this pass — gap. jetson-containers ships an `llama_cpp` package with CUDA-enabled builds tagged per L4T line; exact version matrix not verified.", "researched": false, "sources": [] }, { "runtime": "ollama", "jetpack": null, "jetpack_min": null, "version": null, "install": "", "notes": "Not independently researched this pass — gap. Community reports of running upstream ollama (not Jetson-specific) with CUDA via jetson-containers or manual CUDA toolkit install; no NVIDIA-published compatibility matrix found.", "researched": false, "sources": [] }, { "runtime": "mlc", "jetpack": null, "jetpack_min": null, "version": null, "install": "", "notes": "Not independently researched this pass — gap. Referenced in JetPack 7.0 marketing copy as one of several 'Jetson AI Lab' frameworks (MLC, llama.cpp, Ollama, Hugging Face Transformers) supported on Thor, without version specifics.", "researched": false, "sources": [ { "publisher": "NVIDIA (marketing mention only, not a version table)", "document": "JetPack 7.0 blog/announcement coverage", "doc_id": "forums.developer.nvidia.com/t/343127", "url": "https://forums.developer.nvidia.com/t/jetpack-7-0-jetson-linux-38-2-for-nvidia-jetson-thor-is-now-live/343127", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "C", "quote": "Support for AI serving frameworks includes vLLM, SGLang, and support for frameworks from Jetson AI Lab such as MLC, LAMA CPP, OLAMA, and Hugging Face" } ] } ], "known_issues": { "iso_no_super_upgrade": { "id": "iso_no_super_upgrade", "modules": [ "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb" ], "jetpack": "6.2+ (still true as of 7.2)", "symptom": "25W and MAXN_SUPER power modes not offered by nvpmodel after installing via the ISO image.", "cause": "The ISO image flashes in place without changing the board's flash configuration (TNSPEC/COMPATIBLE_SPEC); it cannot convert a non-Super-flashed device to Super Mode.", "remediation": "Reflash from an x86 host using SDK Manager (or flash.sh) with the jetson-orin-nano-devkit-super / -super-maxn configuration, or manually correct /etc/nv_boot_control.conf's TNSPEC/COMPATIBLE_SPEC to carry the -super suffix.", "sources": [ { "publisher": "NVIDIA", "document": "NVIDIA Developer Forums — 25W and MAXN SUPER not seen in JetPack 7.2 (staff reply, DavidDDD)", "doc_id": "forums.developer.nvidia.com/t/372627", "url": "https://forums.developer.nvidia.com/t/25w-and-maxn-super-not-seen-in-jetpack-7-2/372627", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Currently, the ISO image cannot upgrade a device from non-Super Mode to Super Mode." } ] }, "jp5_to_jp6_no_ota": { "id": "jp5_to_jp6_no_ota", "modules": [ "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial", "jetson_xavier_nx", "jetson_agx_xavier" ], "jetpack": "5.x → 6.x", "symptom": "`apt upgrade`/OTA cannot move a device from JetPack 5.x to 6.x; the device stays on 5.x or breaks partway through.", "cause": "JetPack 5 and 6 use different L4T major versions (R35 vs R36), different base Ubuntu (20.04 vs 22.04), and a different bootloader/kernel — no in-place upgrade path exists across major L4T versions.", "remediation": "Reflash the device fully via SDK Manager or flash.sh with the JetPack 6.x BSP; only same-major-version point-release OTA (e.g. 6.0→6.2.x) is supported.", "sources": [ { "publisher": "NVIDIA", "document": "NVIDIA Developer Forums — Expected JetPack 5 to JetPack 6 Upgrade paths (staff reply, linuxdev)", "doc_id": "forums.developer.nvidia.com/t/280782", "url": "https://forums.developer.nvidia.com/t/expected-jetpack-5-to-jetpack-6-upgrade-paths/280782", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Correct, you won't be able to OTA update from 5 to 6. You'll be able to upgrade minor upgrades only. You'll need to flash (in" } ] }, "jp72_orin_nano_firmware_prereq": { "id": "jp72_orin_nano_firmware_prereq", "modules": [ "jetson_orin_nano", "jetson_orin_nano_4gb" ], "jetpack": "6.x and 7.2.1", "symptom": "Orin Nano Developer Kit fails to boot a freshly flashed JetPack 6.x or 7.2.1 image; hangs or fails at boot.", "cause": "Some Orin Nano devkits shipped with factory UEFI/QSPI firmware older than the JetPack-6.x-generation firmware baseline (pre-36.0); that older firmware cannot boot a JetPack 6.x/7.2.1 image.", "remediation": "Check firmware version; if older than 36.0, run the JetPack 6.x QSPI/UEFI firmware update path (via a JetPack 5.1.3 bridge image, per NVIDIA's documented procedure) before flashing the target JetPack release.", "sources": [ { "publisher": "NVIDIA", "document": "Jetson Orin Nano Developer Kit User Guide — Update Firmware", "doc_id": "nvidia-docs:jetson/orin-nano-devkit/user-guide/latest/update_firmware.html", "url": "https://docs.nvidia.com/jetson/orin-nano-devkit/user-guide/latest/update_firmware.html", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Some developer kits shipped with factory firmware that cannot boot JetPack 6.x or JetPack 7.2.1. If the firmware version is older than 36.0, update the" } ] }, "no_sd_card_jp72": { "id": "no_sd_card_jp72", "modules": [ "jetson_orin_nano" ], "jetpack": "7.2+", "symptom": "No SD Card image download option for the Orin Nano Developer Kit starting with JetPack 7.2.", "cause": "NVIDIA discontinued the SD Card image format for the Orin Nano Developer Kit in favor of a single unified ISO image flashed via USB stick.", "remediation": "Use the unified ISO image and a USB stick to flash the Orin Nano Developer Kit target-side; SDK Manager/flash.sh from an x86 host remains available as an alternative.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack SDK Downloads and Notes Archive — 7.2", "doc_id": "nvidia-dev:embedded/jetpack/downloads/archive-7.2", "url": "https://developer.nvidia.com/embedded/jetpack/downloads/archive-7.2", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "There is no more SD Card image for Jetson Orin Nano Developer Kit. Use the unified ISO image to flash the developer kit using a" } ] }, "thor_requires_jetpack7": { "id": "thor_requires_jetpack7", "modules": [ "jetson_thor_t5000", "jetson_thor_t4000" ], "jetpack": "7.0 minimum for T5000; 7.1 minimum for T4000", "symptom": "Jetson Thor modules cannot be flashed with any JetPack 5.x/6.x image.", "cause": "Thor is a new SoC generation (Blackwell-class GPU, SBSA-aligned firmware/bootloader) with its own L4T line (R38+) starting at JetPack 7.0; T4000 specifically was not supported until JetPack 7.1 (Jetson Linux 38.4).", "remediation": "Flash Thor T5000 with JetPack 7.0 or later; flash T4000 with JetPack 7.1 or later.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack SDK Downloads and Notes Archive — 7.0 and 7.1", "doc_id": "nvidia-dev:embedded/jetpack/downloads/archive-7.1", "url": "https://developer.nvidia.com/embedded/jetpack/downloads/archive-7.1", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Jetson Linux 38.4 ... Jetson AGX Thor Developer Kit, Jetson T5000, Jetson T4000." } ] }, "cuda13_thor_only": { "id": "cuda13_thor_only", "modules": [ "jetson_thor_t5000", "jetson_thor_t4000" ], "jetpack": "7.0+", "symptom": "A container or wheel built against CUDA 12.x (the Orin-generation stack) fails to run or link on a Thor device, or vice versa.", "cause": "JetPack 7.x on Thor ships CUDA 13.x (13.0.0 at 7.0/7.1, 13.2.1 at 7.2/7.2.1) while the Orin-generation JetPack 6.x line tops out at CUDA 12.6.10 — the two SoC generations are not binary-compatible at the CUDA toolkit level.", "remediation": "Build/select containers and wheels tagged for the correct Jetson family (Thor vs. Orin) and JetPack major version; do not mix a JetPack-6-built CUDA 12.x artifact onto a JetPack-7/Thor target or vice versa.", "sources": [ { "publisher": "NVIDIA", "document": "JetPack SDK Downloads and Notes Archive — 7.0", "doc_id": "nvidia-dev:embedded/jetpack/downloads/archive-7.0", "url": "https://developer.nvidia.com/embedded/jetpack/downloads/archive-7.0", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "The latest AI compute stack includes CUDA 13, cuDNN 9.12, and TensorRT 10.13." } ] }, "ubuntu2404_jetpack7": { "id": "ubuntu2404_jetpack7", "modules": [ "jetson_thor_t5000", "jetson_thor_t4000", "jetson_orin_nano", "jetson_orin_nano_4gb", "jetson_orin_nx", "jetson_orin_nx_8gb", "jetson_agx_orin_32gb", "jetson_agx_orin", "jetson_agx_orin_industrial" ], "jetpack": "7.x", "symptom": "apt packages, PPAs, or container base images built for Ubuntu 22.04 (JetPack 6.x) fail to install or behave differently on JetPack 7.x.", "cause": "JetPack 7.x moved the root filesystem to Ubuntu 24.04 LTS with Linux Kernel 6.8, up from Ubuntu 22.04 / Kernel 5.15 on JetPack 6.x.", "remediation": "Rebuild or re-pull container images and packages targeting Ubuntu 24.04 / L4T R38+ when moving a workload from a JetPack-6 (Orin) deployment to JetPack-7 (Thor or unified Orin+Thor 7.2+).", "sources": [ { "publisher": "NVIDIA", "document": "JetPack SDK Downloads and Notes Archive — 7.0", "doc_id": "nvidia-dev:embedded/jetpack/downloads/archive-7.0", "url": "https://developer.nvidia.com/embedded/jetpack/downloads/archive-7.0", "retrieved": "2026-09-07", "verified": "2026-09-07", "class": "A", "quote": "Jetson Linux 38.2 ... Ubuntu 24.04 LTS ... L4T Ubuntu 24.04." } ] } }, "gaps": [ "JetPack 5.1.6 and 5.1.7 compute-stack versions (CUDA/TensorRT/cuDNN/VPI) not verified — only L4T mapping confirmed from the JetPack Archive listing page.", "JetPack 6.2.2 and 6.2.3 compute-stack versions not verified beyond the L4T mapping already present in edgeaistack-api/src/v2/data/power_modes_data.js JETPACK_L4T.", "JetPack 5.1.5 has two L4T revisions listed across sources (35.6.1 and 35.6.2); which is canonical/latest for the 5.1.5 label is unresolved.", "DeepStream version mapping is only confirmed for JetPack 6.1 (DeepStream 7.1, official) and inferred/unconfirmed for 6.2, 6.2.1, 7.0, 7.1, 7.2 — no NVIDIA-published DeepStream↔JetPack 7.x compatibility matrix was found; forum threads (e.g. 'DeepStream 8.0/7.1 compatibility issue on Jetson Orin Nano Super with JetPack 7.2') suggest active incompatibility as of this writing.", "No official JetPack release date found for 5.1.3, 5.1.6, 5.1.7, 6.2.2, or 6.2.3 (left null).", "onnxruntime, llama_cpp, ollama, and mlc runtime/version/install specifics were not independently researched this pass — flagged with jetpack:null in runtime_compat and should be a follow-up topic.", "`ls /dev/nvhost*` expected output and the ollama/llama.cpp CUDA-detection verify commands have no NVIDIA-published canonical reference; recorded as best-effort community patterns or left as explicit gaps.", "Jetson T3000 is mentioned in JetPack 7.2.1 release material only as 'emulation' support, not a shipping module — deliberately excluded from module_ids per the task's module list; a real T3000 launch (if any) is a gap for future research.", "Legacy modules (jetson_xavier_nx, jetson_agx_xavier, jetson_nano) 'last supported JetPack' facts: Xavier NX and AGX Xavier last appear in JetPack 5.1.7 (L4T 35.6.5) per the archive listing; Jetson Nano's last JetPack (4.6.6, L4T 32.7.6) was not independently re-verified this pass (carried over from general JetPack Archive listing, 4.x series, not deep-checked against a dedicated Nano EOL notice)." ] }