Hailo-8/8L AI HAT+ AI Acceleration Module for Raspberry Pi 5, 13/26 TOPS, M.2, PCIe Gen3

Overview

The Hailo-8/8L AI HAT+ is an AI acceleration module (HAT+ expansion board) designed for Raspberry Pi 5, providing up to 26 TOPS (Hailo-8) or 13 TOPS (Hailo-8L) edge AI compute to speed up applications such as image recognition, object detection, and natural language processing. It uses an M.2 interface module and is compliant with Raspberry Pi HAT+ specifications, with software support integrated into Raspberry Pis camera software stack.

Key Features

• Two options: 13 TOPS (Hailo-8L) or 26 TOPS (Hailo-8) AI processor
• Enables real-time, low-latency, efficient AI inference on edge devices
• Standard M.2 interface module; PCIe interface form factor compatible with Raspberry Pi 5 PCIe Gen3 standard
• Supported frameworks: TensorFlow, TensorFlow Lite, ONNX, Keras, PyTorch
• Raspberry Pi OS support
• Lower power consumption (typical power consumption: 2.5W)
• Single-chip solution; internal DRAM, no external plug-in DRAM required
• Scalable: supports multi-data stream, multi-neural network, and multi-chip stacking
• Temperature statements provided: operating temperature 0C to 50C; also states extended temperature range -40C to 85C; industrial -40C to 85C; automotive -40C to 105C

For pre-sales questions, compatibility checks, or after-sales support, contact https://rcdrone.top/ or email [email protected].

Specifications

What’s Included

• Hailo-8L AI HAT+ (Raspberry Pi 5 AI HAT+ 13 TOPS expansion board) or Hailo-8 AI HAT+ (Raspberry Pi 5 AI HAT+ 26 TOPS expansion board) (variant-dependent)
• Pin header
• PCIe cable
• Copper column screw

Applications

• Visual image processing: object detection, camera post-processing, robotics, moderate neural workloads
• Examples demonstrated: object detection, image segmentation, posture estimation, segmented applications, monocular depth
• Example sectors listed: process control, security, home automation, robotics; ITS/perimeter protection/access control; ADAS/autonomous driving/in-cabin monitoring; AMR/AGV, AOI, building management robotics, machinery; point of sale, supply chain management

Performance Test Results (as stated)

• YOLOv5 case test: Raspberry Pi 5 = FPS 8; Raspberry Pi 5 + Hailo-8/Hailo-8L = FPS 30
• Human skeleton test: Raspberry Pi 5 = FPS 1; Raspberry Pi 5 + Hailo-8L = FPS 29.88; Raspberry Pi 5 + Hailo-8 = FPS 30.63
• Background segmentation test: Raspberry Pi 5 = FPS 0 (Can’t run); Raspberry Pi 5 + Hailo-8L = FPS 19.9; Raspberry Pi 5 + Hailo-8 = FPS 27.60
• Image detection test: Raspberry Pi 5 = FPS 1; Raspberry Pi 5 + Hailo-8L = FPS 50.6; Raspberry Pi 5 + Hailo-8 = FPS 60.22

Cooling Note

For improved ventilation under high-load computing, use with an active radiator/active cooler to help avoid overheating and maintain full chip performance. The shipping list does not include Raspberry Pi 5 and the active radiator/active cooler.

Details

Add up to 13 TOPS (Hailo-8L) or 26 TOPS (Hailo-8) of edge AI inference to Raspberry Pi 5 in a HAT+ form factor.

Choose between the 13 TOPS and 26 TOPS variants; each kit includes the HAT+, pin header, PCIe cable, and mounting hardware.

Designed for real-time, low-latency AI inference with a standard M.2 module over Raspberry Pi 5 PCIe Gen3.

A quick comparison chart helps confirm which AI HAT+ option matches your performance and software needs.

Performance examples illustrate higher FPS for common vision workloads like detection and pose estimation when using the Hailo accelerator.

For sustained loads, pairing the setup with the official Raspberry Pi 5 active cooler is recommended for improved airflow.

The Hailo NPU delivers efficient edge inference with typical power around 2.5W and no external DRAM required.

Mechanical dimensions are provided to help plan fitment in enclosures and stacked Raspberry Pi builds.