← All hardware
The first edge accelerator built to run LLMs and VLMs on-device.
Pros
- Brings true generative AI to low-power edge devices
- Direct DDR lets it scale beyond on-chip-only accelerators
- Strong efficiency vs GPU-based alternatives
Cons
- Newer product with a smaller ecosystem than NVIDIA
- 40 TOPS still modest for very large models
- Requires a capable host and memory configuration
✓ Where it shines / best for
- Local LLM/VLM chat assistants on edge hardware
- Privacy-preserving generative AI in IoT and automotive
- Adding generative capability to Raspberry Pi 5 (via AI HAT+ 2)
✕ Not the best fit for
- Heavy multimodal robotics needing 100+ TOPS
- Pure low-cost CNN vision (Hailo-8 is cheaper)
- Datacenter-scale model serving
Features
- ✓ 40 INT4 TOPS generative-AI performance
- ✓ Direct DDR interface to scale to large models
- ✓ Runs LLMs, VLMs and Stable Diffusion locally
- ✓ M.2 form factor with PCIe interface
- ✓ High power efficiency (sub-5W envelope) for always-on edge use
- ✓ Hailo software suite and Model Zoo support
- ✓ Backward-compatible tooling with Hailo-8
- ✓ Targets automotive, IoT and personal compute
Pricing
| Plan | Price | Billing | Notes |
|---|---|---|---|
| Hailo-10H M.2 Module (40 TOPS INT4) | Contact / via distributors | one-time | No public first-party price; small quantities via Hailo global distributors, bulk via volume quote. Appears in retail products such as the $130 RPi AI HAT+ 2 |
Pricing verified from the official source. Prices change often — confirm on the vendor's site before buying.
Specifications
| power | Low-power edge (sub-5W, efficiency-optimized) |
| memory | External DDR (configurable) |
| interface | PCIe (M.2) |
| architecture | Hailo-10H neural processor with direct DDR interface |
| ai_performance | 40 INT4 TOPS |
Sponsored
A full review is being generated for this product and will appear here shortly.