{"product_id":"ai-hat-plus-raspberry-pi-5","title":"Sixfab AI HAT+ for Raspberry Pi 5","description":"\u003cdiv class=\"sxd\"\u003e\n\u003cdiv class=\"sxd-intro\"\u003e\n\u003cp class=\"sxd-lead\"\u003eBuild and test vision AI models directly on Raspberry Pi 5. Sixfab AI HAT+ delivers compact DEEPX AI acceleration for rapid prototyping and local inference.\u003c\/p\u003e\n\u003cp\u003eRun vision AI workloads on Raspberry Pi 5 in real time. Locally, no cloud, no GPU. Plug in the HAT+, install one APT package, and ship inference on your own hardware. As an Official Raspberry Pi Design Partner, Sixfab integrates the DEEPX DX-M1M family directly onto a HAT+ compliant board, giving Pi 5 developers production-grade NPU acceleration over native PCIe without leaving the Raspberry Pi ecosystem.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-stats\"\u003e\n\u003ch3 class=\"sxd-vh\"\u003ePerformance at a glance\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cstrong\u003eUp to 25 TOPS\u003c\/strong\u003e\u003cspan\u003eINT8 · DEEPX DX-M1M\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003ePCIe Gen 3 ×1\u003c\/strong\u003e\u003cspan\u003eNative Pi 5 PCIe\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003e~3 W\u003c\/strong\u003e\u003cspan\u003eTypical NPU draw\u003c\/span\u003e\n\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eHighlights\u003c\/h3\u003e\n\u003cul class=\"sxd-dl\"\u003e\n\u003cli\u003e\n\u003cstrong\u003e25 TOPS at INT8\u003c\/strong\u003e\u003cspan\u003eDEEPX DX-M1M, 2 GB LPDDR4X · 13 TOPS variant with DX-M1ML.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003ePCIe Gen 3 ×1\u003c\/strong\u003e\u003cspan\u003eNative Pi 5 PCIe via 16-pin FFC cable. No USB hops, no bandwidth bottleneck.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003e~3 W typical\u003c\/strong\u003e\u003cspan\u003e3 W typical NPU draw · ~13-15 W combined Pi 5 + HAT+ under load on the official 27 W PSU.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eHAT+ spec compliant\u003c\/strong\u003e\u003cspan\u003eRaspberry Pi HAT+ EEPROM auto-config · 56.5 × 65 mm · stacking-friendly.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eSoldered NPU\u003c\/strong\u003e\u003cspan\u003eSoldered DEEPX silicon transfers heat into the PCB far more efficiently than a socketed M.2 card. No socket to fail, no module slop, no third-party variability.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eAPT install\u003c\/strong\u003e\u003cspan\u003eSigned Sixfab repository ships dxrt-runtime, kernel driver, and tools. Update with apt update.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eDXNN SDK\u003c\/strong\u003e\u003cspan\u003eBring ONNX models from PyTorch, TensorFlow, or Keras. Compile with DX-COM. Deploy with the C++ or Python runtime.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eTechnical specifications\u003c\/h3\u003e\n\u003ctable class=\"sxd-spec\"\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eAI accelerator\u003c\/th\u003e\n\u003ctd\u003eDEEPX DX-M1M (25 TOPS) or DX-M1ML (13 TOPS) at INT8\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eNPU memory\u003c\/th\u003e\n\u003ctd\u003e2 GB LPDDR4X (DX-M1M) · 1 GB LPDDR4X (DX-M1ML)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eHost interface\u003c\/th\u003e\n\u003ctd\u003ePCIe Gen 3 ×1 over 16-pin FFC cable\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eForm factor\u003c\/th\u003e\n\u003ctd\u003eRaspberry Pi HAT+ · 56.5 × 65 mm · 6.56 mm tall\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003ePower input\u003c\/th\u003e\n\u003ctd\u003e5 V via the Pi 5 2×20-pin header and PCIe connector\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eNPU power draw\u003c\/th\u003e\n\u003ctd\u003e3 W (typical)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eCooling\u003c\/th\u003e\n\u003ctd\u003ePassive\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eOperating temp\u003c\/th\u003e\n\u003ctd\u003e0-70 °C commercial\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eSupported host\u003c\/th\u003e\n\u003ctd\u003eRaspberry Pi 5\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eHost OS\u003c\/th\u003e\n\u003ctd\u003eRaspberry Pi OS (Trixie)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eRuntime\u003c\/th\u003e\n\u003ctd\u003edxrt-runtime · APT install · Python \u0026amp; C++ APIs\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eModel pipeline\u003c\/th\u003e\n\u003ctd\u003eONNX → DXNN via DX-COM compiler\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eHot-plug\u003c\/th\u003e\n\u003ctd\u003eNot supported. Power off Pi 5 before mounting\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr class=\"sxd-group\"\u003e\u003cth colspan=\"2\"\u003eCompliance\u003c\/th\u003e\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eCertification\u003c\/th\u003e\n\u003ctd\u003eCertification in progress: CE, FCC, UKCA, RoHS, REACH\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eTwo SKUs · one board\u003c\/h3\u003e\n\u003cp\u003eSame PCB, same HAT+ form factor, same software stack. The NPU module is the only difference. Pick the variant that fits your workload and budget.\u003c\/p\u003e\n\u003cdiv class=\"sxd-scroll\"\u003e\u003ctable class=\"sxd-table\"\u003e\n\u003cthead\u003e\u003ctr\u003e\n\u003ctd\u003e\u003c\/td\u003e\n\u003cth scope=\"col\"\u003eAI HAT+ 13 TOPS\u003c\/th\u003e\n\u003cth scope=\"col\"\u003eAI HAT+ 25 TOPS\u003c\/th\u003e\n\u003c\/tr\u003e\u003c\/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eNPU\u003c\/th\u003e\n\u003ctd\u003eDEEPX DX-M1ML\u003c\/td\u003e\n\u003ctd\u003eDEEPX DX-M1M\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003ePrecision\u003c\/th\u003e\n\u003ctd\u003eINT8\u003c\/td\u003e\n\u003ctd\u003eINT8\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eNPU memory\u003c\/th\u003e\n\u003ctd\u003e1 GB LPDDR4X\u003c\/td\u003e\n\u003ctd\u003e2 GB LPDDR4X\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eBest for\u003c\/th\u003e\n\u003ctd\u003eSingle-camera, single-model\u003c\/td\u003e\n\u003ctd\u003eMulti-model, multi-camera\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eTypical scenario\u003c\/th\u003e\n\u003ctd\u003eLow-power projects, cost-sensitive builds\u003c\/td\u003e\n\u003ctd\u003eHigh-resolution streams, headroom for growth\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\u003c\/div\u003e\n\u003cp class=\"sxd-note\"\u003eAvailable by Q4, 2026.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eHow it works\u003c\/h3\u003e\n\u003col class=\"sxd-steps\"\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-step-n\"\u003e01\u003c\/span\u003e\u003cstrong\u003eCapture\u003c\/strong\u003e\u003cspan\u003eMIPI CSI, USB UVC, or RTSP IP cameras feed frames into the Raspberry Pi 5 - up to 4×1080p.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-step-n\"\u003e02\u003c\/span\u003e\u003cstrong\u003eHost\u003c\/strong\u003e\u003cspan\u003ePi 5's Cortex-A76 CPU runs your app: pre-processing, control flow, I\/O, network.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-step-n\"\u003e03\u003c\/span\u003e\u003cstrong\u003eInference\u003c\/strong\u003e\u003cspan\u003eDEEPX DX-M1M or DX-M1ML NPU runs your compiled DXNN model over PCIe Gen 3 ×1 - 25 TOPS · INT8.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-step-n\"\u003e04\u003c\/span\u003e\u003cstrong\u003eUse results\u003c\/strong\u003e\u003cspan\u003eDetections, segments, and classifications return to your app. Display, log, trigger, stream - 30-35 FPS.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003c\/ol\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eTwo paths to deployment\u003c\/h3\u003e\n\u003cdiv class=\"sxd-cards\"\u003e\n\u003cdiv class=\"sxd-card\"\u003e\n\u003cp class=\"sxd-card-k\"\u003eOption 1\u003c\/p\u003e\n\u003ch4\u003eSixfab Model Zoo\u003c\/h4\u003e\n\u003cp\u003ePre-compiled DXNN models - ready to run, no training required. YOLOv8n, MobileNet, ResNet, and more, already compiled for the DEEPX NPU.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-card\"\u003e\n\u003cp class=\"sxd-card-k\"\u003eOption 2\u003c\/p\u003e\n\u003ch4\u003eDEEPX DXNN SDK\u003c\/h4\u003e\n\u003cp\u003eFull custom model deployment: export your PyTorch, TensorFlow, or Keras model to ONNX, compile to DXNN with DX-COM, and run it through the Python or C++ runtime. INT8 quantization is automatic, with ~2% accuracy delta vs the FP32 source.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eWhat you can build\u003c\/h3\u003e\n\u003cul class=\"sxd-dl\"\u003e\n\u003cli\u003e\n\u003cstrong\u003eVideo analytics cameras\u003c\/strong\u003e\u003cspan\u003eOn-device object detection, counting, intrusion analytics, and retail insights on a single Pi 5 unit.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eRobotics \u0026amp; autonomous systems\u003c\/strong\u003e\u003cspan\u003eReal-time perception, object tracking, and navigation assistance on AMRs, robot arms, and visual-inspection rigs.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eSmart city \u0026amp; infrastructure\u003c\/strong\u003e\u003cspan\u003eTraffic monitoring, facility management, and safety systems on roadside Pi 5 units.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eIndustrial automation\u003c\/strong\u003e\u003cspan\u003eDefect detection, quality inspection, and process monitoring on the production floor.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eDrones \u0026amp; autonomous systems\u003c\/strong\u003e\u003cspan\u003eOn-board perception with low weight and ~3 W typical NPU draw.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eEdge servers \u0026amp; AIoT\u003c\/strong\u003e\u003cspan\u003eCompact inference nodes for multi-camera deployments.\u003c\/span\u003e\n\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eSixfab edge AI ecosystem\u003c\/h3\u003e\n\u003cp\u003eOne NPU, one SDK, three form factors for real-world deployment.\u003c\/p\u003e\n\u003cul class=\"sxd-rows\"\u003e\n\u003cli\u003e\n\u003cdiv class=\"sxd-row-h\"\u003e\n\u003cstrong\u003eAI HAT+\u003c\/strong\u003e\u003cspan\u003eThis product\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cp\u003e13 \/ 25 TOPS AI acceleration for Raspberry Pi 5.\u003c\/p\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cdiv class=\"sxd-row-h\"\u003e\n\u003ca href=\"\/products\/edge-ai-expansion-board-raspberry-pi-5\"\u003eSixfab Edge AI Expansion Board\u003c\/a\u003e\u003cspan\u003e25 TOPS · LTE\/5G\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cp\u003eAdd LTE\/5G connectivity, NVMe local storage, and multi-camera support in a single under-board stack.\u003c\/p\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cdiv class=\"sxd-row-h\"\u003e\n\u003ca href=\"\/products\/alpon-x5-ai\"\u003eALPON X5 AI\u003c\/a\u003e\u003cspan\u003e25 TOPS · −20 to +60 °C\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cp\u003eFanless, rugged, always-online edge AI computer for fleets and distributed industrial sites.\u003c\/p\u003e\n\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eWhat's in the box\u003c\/h3\u003e\n\u003cul class=\"sxd-box\"\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-box-n\"\u003eSixfab AI HAT+ board\u003cspan class=\"sxd-box-s\"\u003e13 TOPS · INT8 (DEEPX DX-M1ML) or 25 TOPS · INT8 (DEEPX DX-M1M), depending on the option\u003c\/span\u003e\u003c\/span\u003e\u003cspan class=\"sxd-box-q\"\u003e×1\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-box-n\"\u003ePCIe FFC cable (16-pin)\u003c\/span\u003e\u003cspan class=\"sxd-box-q\"\u003e×1\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-box-n\"\u003e16 mm stacking header (2×20, 2.54 mm)\u003c\/span\u003e\u003cspan class=\"sxd-box-q\"\u003e×1\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-box-n\"\u003eM2.5 × 16 mm F-F spacer\u003c\/span\u003e\u003cspan class=\"sxd-box-q\"\u003e×4\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-box-n\"\u003eM2.5 × 5 mm plastic screw\u003c\/span\u003e\u003cspan class=\"sxd-box-q\"\u003e×8\u003c\/span\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003cspan class=\"sxd-box-n\"\u003ePassive cooler (with thermal pad)\u003c\/span\u003e\u003cspan class=\"sxd-box-q\"\u003e×1\u003c\/span\u003e\n\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003cp class=\"sxd-fine\"\u003eBoth options include the exact same 6-item mounting kit and assembly hardware; the only difference is which DEEPX NPU is soldered on the board. For Raspberry Pi 5 only.\u003c\/p\u003e\n\u003ch4 class=\"sxd-sub\"\u003eNot included (sold separately)\u003c\/h4\u003e\n\u003ctable class=\"sxd-kv sxd-ni\"\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eRequired\u003c\/th\u003e\n\u003ctd\u003eRaspberry Pi 5 (host board) · official 27 W USB-C PD power supply for the Raspberry Pi 5 · microSD card flashed with Raspberry Pi OS\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eRecommended\u003c\/th\u003e\n\u003ctd\u003eRaspberry Pi Active Cooler for the Pi 5 (for sustained 100% NPU utilization)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eOptional\u003c\/th\u003e\n\u003ctd\u003eUSB or CSI camera (for live vision inference workloads)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eCompatibility\u003c\/h3\u003e\n\u003cdiv class=\"sxd-host\"\u003e\n\u003cdiv class=\"sxd-host-yes\"\u003e\n\u003cp class=\"sxd-k\"\u003eSupported host\u003c\/p\u003e\n\u003cul\u003e\u003cli\u003eRaspberry Pi 5\u003c\/li\u003e\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-host-no\"\u003e\n\u003cp class=\"sxd-k\"\u003eNot supported\u003c\/p\u003e\n\u003cul\u003e\n\u003cli\u003eRaspberry Pi 4\u003c\/li\u003e\n\u003cli\u003eCompute Module 4\u003c\/li\u003e\n\u003cli\u003eCompute Module 5 on the CM5 IO Board\u003c\/li\u003e\n\u003cli\u003eNon-Raspberry Pi SBCs\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003ctable class=\"sxd-kv sxd-compat\"\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eCameras\u003c\/th\u003e\n\u003ctd\u003e\n\u003cspan class=\"sxd-item\"\u003eRaspberry Pi Camera Modules \u003cspan class=\"sxd-muted\"\u003e(MIPI CSI)\u003c\/span\u003e\u003c\/span\u003e\u003cspan class=\"sxd-sep\"\u003e · \u003c\/span\u003e\u003cspan class=\"sxd-item\"\u003eUSB cameras \u003cspan class=\"sxd-muted\"\u003e(UVC)\u003c\/span\u003e\u003c\/span\u003e\u003cspan class=\"sxd-sep\"\u003e · \u003c\/span\u003e\u003cspan class=\"sxd-item\"\u003eIP cameras \u003cspan class=\"sxd-muted\"\u003e(RTSP)\u003c\/span\u003e\u003c\/span\u003e\u003cspan class=\"sxd-sep\"\u003e · \u003c\/span\u003e\u003cspan class=\"sxd-item\"\u003eMulti-camera configurations\u003c\/span\u003e\n\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eFrameworks\u003c\/th\u003e\n\u003ctd\u003e\n\u003cspan class=\"sxd-item\"\u003eONNX \u003cspan class=\"sxd-muted\"\u003e(primary)\u003c\/span\u003e\u003c\/span\u003e\u003cspan class=\"sxd-sep\"\u003e · \u003c\/span\u003e\u003cspan class=\"sxd-item\"\u003ePyTorch\u003c\/span\u003e\u003cspan class=\"sxd-sep\"\u003e · \u003c\/span\u003e\u003cspan class=\"sxd-item\"\u003eTensorFlow\u003c\/span\u003e\u003cspan class=\"sxd-sep\"\u003e · \u003c\/span\u003e\u003cspan class=\"sxd-item\"\u003eKeras\u003c\/span\u003e\u003cspan class=\"sxd-sep\"\u003e · \u003c\/span\u003e\u003cspan class=\"sxd-item\"\u003eUltralytics YOLO \u003cspan class=\"sxd-muted\"\u003e(native integration coming soon)\u003c\/span\u003e\u003c\/span\u003e\n\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003c\/div\u003e\n\u003cdiv class=\"sxd-sec\"\u003e\n\u003ch3\u003eFrequently asked questions\u003c\/h3\u003e\n\u003cdiv class=\"sxd-faq\"\u003e\n\u003cdetails open\u003e\u003csummary\u003eWhat's the difference between 13 TOPS and 25 TOPS?\u003c\/summary\u003e\u003cp\u003e13 TOPS (DEEPX DX-M1ML): single-model, single-camera deployments. Low power. 25 TOPS (DEEPX DX-M1M): multi-model pipelines, multi-camera, high-resolution. Recommended for most projects. Both share the same PCB, HAT+ form factor, software stack, and DXNN SDK. Only the soldered NPU module differs.\u003c\/p\u003e\u003c\/details\u003e\n\u003cdetails\u003e\u003csummary\u003eHow do I decide whether 13 or 25 TOPS is right for my project?\u003c\/summary\u003e\u003cp\u003eSize it from three things: how many camera streams and models you run at once, your target input resolution, and whether you want headroom for future models. Pick 13 TOPS (DX-M1ML) for single-camera, single-model pipelines, prototypes, and budget-sensitive builds at common resolutions. Pick 25 TOPS (DX-M1M) for multi-stream 1080p, higher-resolution single-stream, multi-model pipelines, or larger compiled models - it is the safe default when you want room to grow. To estimate before you buy, check the per-model FPS figures in the Sixfab Model Zoo against your target frame rate and stream count.\u003c\/p\u003e\u003c\/details\u003e\n\u003cdetails\u003e\u003csummary\u003eHow does it compare to the Raspberry Pi AI HAT+ and AI HAT+ 2?\u003c\/summary\u003e\u003cp\u003eAll three are HAT+ form-factor accelerators that mount on a Raspberry Pi 5 and run inference on a soldered NPU over PCIe. The difference is the silicon: Sixfab AI HAT+ uses DEEPX (DX-M1ML at 13 TOPS or DX-M1M at 25 TOPS, INT8); the Raspberry Pi boards use Hailo. The Raspberry Pi AI HAT+ ships as Hailo-8L (13 TOPS) or Hailo-8 (26 TOPS); for vision, Sixfab's 25 TOPS at INT8 is competitive with the 26 TOPS Hailo-8 model, and YOLOv8n at 640×640 runs 30-35 FPS on a Raspberry Pi 5 with 8 GB RAM. The newer Raspberry Pi AI HAT+ 2 (Hailo-10H, 40 TOPS at INT4, 8 GB on-board RAM) adds on-device LLMs and VLMs; the Sixfab AI HAT+ is vision-focused today. If your project specifically needs local generative AI right now, the Raspberry Pi AI HAT+ 2 is built for that workload.\u003c\/p\u003e\u003c\/details\u003e\n\u003cdetails\u003e\u003csummary\u003eCan I run LLMs on this?\u003c\/summary\u003e\u003cp\u003eNo. DEEPX DX-M1M and DX-M1ML are optimized for computer vision (object detection, segmentation, classification). The current generation doesn't support LLMs. LLMs are on the DEEPX roadmap and Sixfab will support them as the silicon enables.\u003c\/p\u003e\u003c\/details\u003e\n\u003cdetails\u003e\u003csummary\u003eHow long does setup take?\u003c\/summary\u003e\u003cp\u003eUnder 15 minutes: power off the Pi 5, mount the AI HAT+, connect the 16-pin FFC cable, install the dxrt-runtime APT package, verify with lspci | grep DEEPX, run a Model Zoo demo, and see YOLOv8n at 640×640 hit 30-35 FPS on Raspberry Pi 5 with 8 GB RAM. Custom DXNN SDK deployments take 1-2 hours for ONNX export, DXNN compilation, and application integration.\u003c\/p\u003e\u003c\/details\u003e\n\u003cdetails\u003e\u003csummary\u003eDoes it work offline?\u003c\/summary\u003e\u003cp\u003eYes, completely. Inference runs entirely on-device over the PCIe Gen 3 x1 link between the Raspberry Pi 5 and the DEEPX NPU. No cloud, no GPU, no external connectivity required.\u003c\/p\u003e\u003c\/details\u003e\n\u003cdetails\u003e\u003csummary\u003eWhich Raspberry Pi models are supported?\u003c\/summary\u003e\u003cp\u003eRaspberry Pi 5 is the only supported host. Not supported: Raspberry Pi 4, Compute Module 4, the Raspberry Pi Compute Module 5 on the Raspberry Pi CM5 IO Board, and non-Raspberry Pi SBCs. Hot-plug is not supported - power off the Raspberry Pi 5 before mounting or removing the AI HAT+.\u003c\/p\u003e\u003c\/details\u003e\n\u003cdetails\u003e\u003csummary\u003eWhat camera formats are supported?\u003c\/summary\u003e\u003cp\u003eRaspberry Pi Camera Modules (MIPI CSI), USB cameras (UVC), IP cameras (RTSP), and multi-camera configurations. Cameras connect directly to the Raspberry Pi 5; the AI HAT+ does not obstruct the Pi 5's CSI connectors.\u003c\/p\u003e\u003c\/details\u003e\n\u003cdetails\u003e\u003csummary\u003eWhat AI frameworks are supported?\u003c\/summary\u003e\u003cp\u003eONNX (primary), PyTorch, TensorFlow, Keras, and Ultralytics YOLO (native integration coming soon). Models are exported to ONNX, then compiled to DXNN via the DEEPX DXNN SDK for execution on the NPU. The Sixfab Model Zoo includes pre-compiled models (YOLOv8n, YOLOv8s, MobileNet, ResNet, and others) ready to deploy.\u003c\/p\u003e\u003c\/details\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e","brand":"Sixfab","offers":[{"title":"25 TOPS (DX-M1M)","offer_id":48916963295459,"sku":"S166","price":90.0,"currency_code":"EUR","in_stock":true},{"title":"13 TOPS (DX-M1ML)","offer_id":48953913999587,"sku":"N\/A","price":63.0,"currency_code":"EUR","in_stock":false}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/0795\/4618\/8003\/files\/Sixfab-AI-HAT-Plus-11.jpg?v=1781616885","url":"https:\/\/eu.sixfab.com\/products\/ai-hat-plus-raspberry-pi-5","provider":"Sixfab","version":"1.0","type":"link"}