Home Hailo Hailo-8L Entry-Level AI Accelerator
Image 1
Image 2
Hailo

Hailo-8L Entry-Level AI Accelerator

The Hailo-8L AI Accelerator offers 13 TOPS of processing power, designed for entry-level AI applications. It delivers real-time, low-latency AI inferencing on edge devices with high-efficiency processing. With a typical power consumption of just 1.5W, the Hailo-8L ensures best-in-class power efficiency. Its compact design does not require external memory, making it a highly cost-effective solution. Compatible with the Hailo-8 software suite, it also allows easy migration to higher AI capabilities for more demanding applications..

  • Detail
  • Special Features
  • Specification

The Hailo-8L AI Accelerator is designed for edge devices requiring entry-level AI capabilities. With 13 TOPS of processing power, it provides high-efficiency AI inferencing, enabling edge applications to run deep learning models in real time with low latency and high efficiency. The Hailo-8L offers a cost-effective solution for entry-level AI-powered devices, including AI video analytics, image recognition, and object detection, all while consuming a minimal amount of power (only 1.5W).

The accelerator is compatible with major AI frameworks such as TensorFlow, TensorFlow Lite, Keras, PyTorch, and ONNX, and supports both X86 and ARM host architectures, making it easy to integrate into existing systems. Additionally, the Hailo-8L comes with a comprehensive dataflow compiler to ensure quick and easy deployment of neural network models.

With scalable processing capabilities, the Hailo-8L can simultaneously handle multiple streams and models, making it suitable for a wide variety of applications in industries such as security, retail, and industrial automation. The Hailo-8L is ideal for use in devices that require real-time performance at low power consumption and can be upgraded to more powerful systems like the Hailo-8 as future needs evolve.

  • 13 TOPS Performance

    • The Hailo-8L delivers 13 tera-operations per second (TOPS), offering high-performance AI inferencing for entry-level edge devices while maintaining exceptional efficiency.

  • Real-Time, Low-Latency AI Inferencing

    • Provides real-time AI processing with low-latency and high efficiency, enabling the processing of multiple AI tasks and models in parallel, perfect for AI-driven edge devices.

  • Best-in-Class Power Efficiency

    • With a typical power consumption of only 1.5W, the Hailo-8L ensures maximum energy efficiency, making it ideal for edge applications that require both performance and power savings.

  • Compact and Simple Integration

    • The Hailo-8L does not require external memory, simplifying hardware integration and reducing system complexity for a more efficient AI deployment.

  • Scalability

    • Capable of simultaneously processing multiple streams and models, the Hailo-8L supports a wide range of applications from image classification to semantic segmentation.

  • Future-Ready

    • The Hailo-8L is compatible with the Hailo-8 software suite, allowing easy migration to more powerful AI solutions when higher capacity is required in the future.

  • Industrial Grade Reliability

    • Designed to operate in challenging environments, the Hailo-8L supports industrial-grade conditions with an operating temperature range from -40°C to 85°C.

General Features

  • AI Performance: 13 TOPS

  • Typical Power Consumption: 1.5W

  • Latency: Real-time, low-latency AI inferencing

  • External Memory: Not required

  • Operating Temperature: Industrial grade, -40°C to 85°C

  • Power Efficiency: Best-in-class TOPS/$

Supported Frameworks & Architectures

  • AI Frameworks: TensorFlow, TensorFlow Lite, Keras, PyTorch, ONNX

  • Host Architectures: X86, ARM

  • Operating Systems: Linux, Windows

Dataflow Compiler & Compatibility

  • Comprehensive dataflow compiler for easy and quick porting of neural network models

  • Future-ready: Seamless migration to Hailo-8 for higher AI capacities

Benchmarks

NN Model Input Resolution FPS Power (W) FPS/W
ResNet-50 v1 224×224 504 1.9 260
ViT_base 224×224 122 2.5 48
YOLOv5m 640×640 106 2.9 37
YOLOv8m 640×640 87 3.1 28
YOLOv11m 640×640 58 2.3 26
stdc1 1,024×1,920 20 1.7 12

 

Batch 8, measurements taken at room temperature through PCIe interface on Hailo-8L evaluation board. System host: Intel® Core™ i5-9400 CPU @ 2.90GHz; Models compiled with Hailo Dataflow Compiler version 3.31.0.