US Hailo
// Headquartered in Tel Aviv, Israel. Founded in 2017.
Hailo AI Accelerators for the Edge
Hailo designs AI processors for edge devices, with a dataflow architecture that keeps the neural network on the chip instead of moving data to and from external memory. That gives high inference throughput per watt in a module small enough for an M.2 slot, so an x86 or Arm computer can run vision models at a few watts without a GPU. The Hailo-10H adds memory on the module for generative AI, and the same Dataflow Compiler and HailoRT runtime serve every chip.
01 Hailo Accelerators
Hailo AI Accelerators by Chip
Three Hailo accelerators, from the entry Hailo-8L to the Hailo-10H for generative AI, with the series and models we build around each. Pick the chip that fits the model, the camera count and the power budget, then the edge AI computer it goes into.
Hailo-8 Edge AI Computers
The Hailo-8 is the most widely used chip in the range: up to 26 TOPS at a few watts, enough for real-time detection, classification and tracking across several camera streams beside a low-power x86 or Arm processor. It runs our fanless box PCs, rail and outdoor computers and AMD Ryzen industrial PCs.
Suited to
Machine VisionVideo AnalyticsRail CCTVSmart Cities
Hailo-8
26 TOPS Vision Accelerator
- Performance
- Up to 26 TOPS
- Power
- Around 2.5 W typical
- Memory
- None external: the network runs on the chip
- Interface
- PCIe Gen3, up to four lanes
- Form Factor
- M.2 and mini PCIe modules
- Systems
- Fanless box PCs · rail and IP67 outdoor computers · in-vehicle and AMD Ryzen industrial PCs
Series and Models With This GPU
Hailo-8L Entry-Level AI Acceleration
The Hailo-8L is the entry-level version of the Hailo-8, at up to 13 TOPS, for single-camera and lighter vision models where the full Hailo-8 would be more than the application needs. It uses the same software, so a model can move between the two chips.
Suited to
Single-Camera VisionKiosksSmart SensorsPrototyping
Hailo-8L
13 TOPS Entry Accelerator
- Performance
- Up to 13 TOPS
- Power
- Lower than the Hailo-8
- Memory
- None external: the network runs on the chip
- Interface
- PCIe Gen3
- Form Factor
- M.2 modules
- Systems
- Specified to order in systems with a matching M.2 slot
Hailo-10H Generative AI Computers
The Hailo-10H brings generative AI to edge devices: up to 40 TOPS with memory on the module, so small language models, vision language models and speech run locally at a few watts. We fit it in compact AMD Ryzen industrial PCs alongside the processor's own NPU.
Suited to
Generative AIVision Language ModelsVoice InterfacesRobotics
Hailo-10H
Generative AI at the Edge
- Performance
- Up to 40 TOPS (INT4)
- Power
- Around 2.5 W typical
- Memory
- 8 GB LPDDR4 on the module
- Interface
- PCIe Gen3, four lanes
- Form Factor
- M.2 M key module
- Systems
- Compact AMD Ryzen industrial PCs
// LOADING...
Browse Hailo Edge AI Computers
- GPU / AI Accelerator: Hailo-8 AI Accelerator
- CPU: Intel Celeron J Series Elkhart Lake
- AI Performance: 26 TOPS
- Memory: 1x DDR4 3200MHz SO-DIMM Slot, 32GB
- OS: Linux
- Display Outputs: 1x HDMI
- Ethernet: 2x GbE LAN [RJ45]
- USB: 1x USB 2.0 [Type A], 1x USB 3.1 Gen 1 [Type A]
- M.2: 1x M.2 B-Key [PCIe/USB] [3042/3052], 1x M.2 B-Key [USB/SATA] [2242], 1x M.2 E-Key [PCIe x1/USB 2.0] [2230]
- Storage: 1x M.2 SSD
- Power: 12-24V DC [Barrel Connector]
- Certification: CE, FCC
- Mounting Options: Desktop, Wall Mount
- Temp: -10°C to 70°C [14°F to 158°F]
- GPU / AI Accelerator: Hailo-8 AI Accelerator
- CPU: NXP i.MX 8QuadMax
- AI Performance: 26 TOPS
- Memory: 4GB Soldered On-board
- OS: Linux Yocto 3.0
- Display Outputs: 1x HDMI
- Ethernet: 2x GbE LAN [RJ45]
- USB: 2x USB 2.0 [Type A]
- Serial: 2x RS232/422/485 [Terminal Block]
- CANbus: 2x CANbus [Terminal Block]
- M.2: 1x M.2 B-Key [PCIe x2/USB 3.0/SATA] [3052], 1x M.2 M-Key [SATA] [2280]
- mPCIe: 1x mPCIe [PCIe/USB/SIM]
- Storage: 16GB eMMC, 1x M.2 SSD, 1x Micro SD
- Power: 12V DC [Barrel Connector]
- Temp: -20°C to 70°C [-4°F to 158°F]
- Certification: CE, FCC
- Mounting Options: Desktop, Wall Mount
02 Guides
Hailo Edge AI Guides
Longer reads from our engineers on AI accelerators and NPUs, and how they are chosen for industrial edge systems.
Hailo-8 AI Accelerators: Benefits and Integration for Edge Computing
The Hailo-8 runs deep learning inference at up to 26 TOPS on around 2.5 W. This guide covers how it works, how it compares with the Hailo-8L and Hailo-10H, and how to build it into an industrial edge computer.
What Is an NPU? Neural Processing Units for Edge AI
An NPU is a dedicated AI accelerator built into a processor to run machine-learning inference efficiently. It is the hardware behind on-device AI — now across Intel Core Ultra, AMD Ryzen AI and NVIDIA Jetson.
GPU vs NPU: Which for Edge AI?
GPUs and NPUs both accelerate AI, but they are built for different jobs. For an edge-AI design, the choice — or whether you need both — decides the power, size and cost of the system.
M.2 Expansion Slots & Keys: Industrial SSDs, Connectivity, I/O, Accelerators
Quickly becoming the common standard for onboard expansion in embedded and industrial computing, M.2 expansion is replacing the likes of mPCIe and mSATA for expansion and storage capabilities.
// Hailo Edge AI Specialists
Speak to an Edge AI Specialist.
Things Embedded configures, images and tests Hailo-based systems from our partners in the USA. Tell us the model, the cameras and the environment, and we will point you to the right Hailo accelerator and the edge AI computer built around it.
Phone · USA
+1 407 214 9446