Support and discussions for the Sixfab AI HAT+ and Edge AI Expansion Board: DEEPX-powered NPU accelerators (13/25 TOPS) for fast, efficient, on-device AI inference on Raspberry Pi 5.
Welcome to the Raspberry Pi AI HATs category, the central hub for learning, experimenting, and sharing everything related to Sixfab AI acceleration hardware for Raspberry Pi 5. Whether you are exploring edge AI for the first time or developing production-grade vision systems, you will find discussions, guides, and real-world experiences here.
What is an AI HAT for Raspberry Pi 5?
An AI HAT is an add-on board that brings a dedicated NPU (Neural Processing Unit) to the Raspberry Pi 5, letting you run machine learning inference locally with high efficiency and low latency. Instead of sending data to the cloud, your application processes information on-device, which means faster response times, lower bandwidth costs, and improved data privacy.
Both Sixfab boards share the same DEEPX software stack (dxrt-runtime, the DXNN SDK, and the Sixfab Model Zoo), so you can move from prototype to field deployment without rebuilding your pipeline.
Which board should I choose?
| Sixfab AI HAT+ | Sixfab Edge AI Expansion Board | |
|---|---|---|
| Form factor | HAT+ compliant, mounts on top of Pi 5 | Under-board, triple M.2 architecture |
| NPU | Soldered DEEPX DX-M1ML (13 TOPS) or DX-M1M (25 TOPS) at INT8 | DEEPX DX-M1 (25 TOPS at INT8) on a swappable M.2 module |
| Host interface | Native PCIe Gen 3 x1 (16-pin FFC) | PCIe via triple M.2 slots |
| Storage | None | NVMe SSD slot |
| Connectivity | None | LTE/5G modem slot |
| Power | Pi 5 27 W USB-C PD supply | Single USB-C PD, back-powers the Pi 5 |
| Best for | Rapid prototyping, PoCs, low-power single/multi-camera inference | Connected field deployments: local inference + high-speed logging + always-on cellular uplink |
| Product page | sixfab.com/product/ai-hat-plus-raspberry-pi-5 | sixfab.com/product/edge-ai-expansion-board-raspberry-pi-5 |
| Documentation | AI HAT+ docs | Edge AI Expansion Board docs |
What AI HATs enable
- Run AI models locally on Raspberry Pi 5 with low latency
- Process real-time vision and multi-camera streams on-device
- Build privacy-focused edge AI systems with no cloud dependency
- Log data at high speed and stay connected over LTE/5G in the field
- Scale the same models from prototype to production without porting
Topics covered in this category
- Introduction to AI HATs, NPUs, and hardware acceleration on Raspberry Pi 5
- Comparing the AI HAT+ and Edge AI Expansion Board for your use case
- Choosing between the 13 TOPS and 25 TOPS DEEPX variants
- Deploying models with the Sixfab Model Zoo and the DXNN SDK (ONNX to DXNN)
- Improving inference speed, FPS, and system performance
- Working with MIPI CSI, USB (UVC), and IP (RTSP) cameras
- NVMe storage, LTE/5G connectivity, and power or cooling optimization
- Designing edge AI systems for real-world and industrial environments
- Debugging workflows and practical setup tips
What you can build
- Intelligent vision systems and connected smart cameras
- Robotics and autonomous platforms with real-time perception
- Outdoor and remote monitoring with cellular uplink
- Industrial quality control and automation tools
- Edge-based IoT systems with on-device intelligence
- Audio and event detection applications
Join the community to exchange ideas, troubleshoot challenges, and showcase your projects. Whether you are optimizing performance, testing new models, or building something entirely new, this is the place to collaborate and grow.
Products and resources
| Resource | Link |
|---|---|
 Buy AI HAT+ |
sixfab.com/product/ai-hat-plus-raspberry-pi-5 |
| Buy Edge AI Expansion Board |
sixfab.com/product/edge-ai-expansion-board-raspberry-pi-5 |
 Documentation |
AI HAT+ · Edge AI Expansion Board · AI Model Deployment |
 Quickstart guides |
AI HAT+ · Edge AI Expansion Board |
 DEEPX examples on GitHub |
github.com/sixfab/sixfab-dx-examples |
 ALPON Cloud |
connect.sixfab.com |
 All Raspberry Pi HATs |
sixfab.com → Raspberry Pi HATs |