Hailo-8 AI Accelerator

In Stock

Precision You Can Trust
NABL-Accredited Calibration | Available on Request

The Hailo-8 is an AI processor designed for on-device inference in embedded systems. Part number HNC18BI11BH is the industrial processor IC, specified for an operating temperature range of −40°C to 85°C. It delivers up to 26 TOPS of AI compute performance with 2.5W typical processor power consumption.

Request Purchase

Datasheet

Hailo-8 AI Accelerator for embedded systems

The Hailo-8 AI Accelerator enables embedded devices to run neural network inference close to where data is generated. Part number HNC18BI11BH identifies the industrial processor IC, specified for operation from −40°C to 85°C. The processor delivers up to 26 TOPS with 2.5W typical processor power consumption.

How Hailo-8 works

Hailo-8 connects to a compatible x86 or Arm host through PCIe and works as an AI co-processor. Its inference memory is integrated on the processor, so the accelerator does not require external DRAM. The complete device still needs a suitable host, power supply and application software.

The processor’s dataflow architecture supports systems that run multiple AI models or analyse multiple input streams. Actual throughput, latency and power use depend on the selected models, inputs and complete system design.

Applications

Hailo-8 can add local AI processing to smart cameras, video analytics systems, industrial visual inspection equipment, access-control devices and other embedded computer vision products. Processing on the device can help applications act on visual data without sending every input to a cloud service.

Development software

Hailo’s software tools include the Dataflow Compiler for preparing trained models, HailoRT for runtime inference and a Model Zoo with example models. The development workflow accommodates models originating from TensorFlow, TensorFlow Lite, Keras, PyTorch and ONNX. Hailo lists Linux and Windows host support.

HNC18BI11BH is an IC for board-level integration. Hailo-8 M.2 modules and PCIe cards are separate products.

Up to 26 TOPS AI performance: Run neural network inference in an embedded edge device.
2.5W typical processor power: Supports designs with limited power and thermal budgets.
Integrated inference memory: No external DRAM is required for the accelerator IC.
Multi-model and multi-stream support: Build applications with several AI workloads or input streams.
PCIe host connection: Integrate with a compatible x86 or Arm host processor.
Industrial operating range: The HNC18BI11BH IC is specified for −40°C to 85°C.
Software development tools: Prepare and deploy models using the Dataflow Compiler, HailoRT and Model Zoo.
Board-level integration: Suitable for OEMs developing custom embedded AI hardware.

Parameter	Specification
Model	Hailo-8 AI Accelerator
Manufacturer part number	HNC18BI11BH
Product format	Industrial AI accelerator processor IC
AI compute performance	Up to 26 TOPS
Typical processor power consumption	2.5W
Host interface	PCIe
Compatible host architectures	x86 and Arm
Industrial operating temperature	−40°C to 85°C
Accelerator inference memory	Integrated; no external DRAM required
Host operating systems listed by Hailo	Linux and Windows
Development tools	Dataflow Compiler, HailoRT and Model Zoo

FAQs

What is the product number of this Hailo-8 processor?
The industrial processor IC on this page has manufacturer part number HNC18BI11BH.

Is this Hailo-8 product an M.2 module?
No. This listing is for the processor IC used in board-level hardware designs. M.2 modules and PCIe cards are separate products.

Does the Hailo-8 require external memory?
The accelerator has integrated inference memory and does not require external DRAM. Other components in the complete device, including the host processor, may need their own memory.

Can Hailo-8 run multiple AI models?
Yes, it supports multi-model and multi-stream designs. The achievable performance depends on the models and system configuration.