← All hardware
26 TOPS of ruthlessly efficient vision inference at just 2.5W.
Pros
- Outstanding TOPS-per-watt for vision workloads
- Tiny thermal/power footprint enables fanless designs
- Broad host compatibility via M.2/PCIe
Cons
- Optimized for CNN vision, not large generative models
- Needs a host CPU; not a standalone computer
- Toolchain has a learning curve for model porting
✓ Where it shines / best for
- Real-time edge computer vision (security, retail, industrial)
- Embedding high-efficiency CNN inference in M.2-equipped devices
- Multi-camera analytics on edge gateways and NVRs
✕ Not the best fit for
- Generative AI / large LLM workloads (vision/CNN-focused; use Hailo-10H)
- Model training or data-center scale
- Non-technical plug-and-play consumer use
Features
- ✓ Hailo-8 deep learning processor delivering up to 26 TOPS (INT8)
- ✓ Industry-leading efficiency (~3 TOPS/W) in an M.2 module
- ✓ M.2 form factor for embedded systems, NVRs, and industrial PCs
- ✓ Hailo Dataflow Compiler and extensive model zoo
- ✓ Optimized for real-time multi-stream computer-vision inference
- ✓ Low power (~2.5W typical) for fanless edge enclosures
- ✓ Broad framework support (TensorFlow, ONNX, PyTorch via compiler)
Pricing
| Plan | Price | Billing | Notes |
|---|---|---|---|
| Hailo-8 M.2 Module | ~$70 | one-time | M.2 module with Hailo-8 (26 TOPS); price varies by distributor and volume |
| Hailo-8 Starter / Evaluation Kit | Contact vendor | one-time | Dev kits and M.2 eval boards priced separately |
Pricing verified from the official source. Prices change often — confirm on the vendor's site before buying.
Specifications
| power | ~2.5W typical |
| memory | On-chip (no external DRAM) |
| interface | PCIe Gen-3.0 (M.2 / mini-PCIe) |
| architecture | Hailo-8 dataflow neural processor |
| ai_performance | 26 INT8 TOPS |
Sponsored
A full review is being generated for this product and will appear here shortly.