Sixfab AI HAT+ for Raspberry Pi 5 integrates DEEPX DX-M1 AI accelerator

 Sixfab has launched the AI HAT+ for Raspberry Pi 5, a PCIe-based AI accelerator board powered by the DEEPX DX-M1 neural processing unit (NPU). The board is designed to bring high-performance edge AI capabilities to Raspberry Pi 5 systems, enabling developers to run computer vision workloads such as object detection, image classification, and segmentation directly on-device without relying on cloud processing.

The new AI HAT+ follows the Raspberry Pi HAT+ specification and connects to the Raspberry Pi 5 through a PCIe interface using a 16-pin FFC cable. Unlike solutions that rely on M.2 accelerator modules, the DEEPX accelerator is soldered directly onto the board, resulting in a compact design optimized for Raspberry Pi deployments. The AI HAT+ is available in two performance variants delivering up to 13 TOPS and 25 TOPS of AI processing power.

Sixfab AI HAT+ Specifications

Supported SBC

• Raspberry Pi 5

AI Accelerator Options

• DEEPX DX-M1ML
  • Up to 13 TOPS (INT8)
  • 512MB LPDDR4X NPU memory
• DEEPX DX-M1M
  • Up to 25 TOPS (INT8)
  • 1GB LPDDR4X NPU memory

Host Interface

• PCIe Gen 3 x1 via 16-pin FFC cable

Cooling

• Passive heatsink cooling
• 2-pin JST fan connector included

Power

• 5V / 3A supplied through Raspberry Pi 5 GPIO header
• No additional power connector required
• Recommended Raspberry Pi 27W USB-C power supply

Power Consumption

• NPU idle: approximately 0.5W to 1W
• NPU peak load: approximately 2.5W to 3W
• Raspberry Pi 5 + AI HAT+ system: approximately 13W to 15W

Dimensions

• 65 x 56.5 mm
• Raspberry Pi HAT+ compliant form factor

Operating Temperature

• 0°C to 70°C

Certifications

• CE
• FCC
• UKCA
• RoHS
• REACH (in progress)

The AI HAT+ is intended for developers building edge AI applications that require low latency and local processing. By leveraging the DEEPX NPU, AI workloads can be executed directly on the accelerator while freeing the Raspberry Pi's CPU for other tasks. This makes the platform suitable for smart cameras, robotics, industrial monitoring, security systems, and intelligent IoT gateways.

Software Support and AI Models

The board supports Raspberry Pi OS and utilizes Sixfab's software ecosystem, including the DXRT runtime environment and DXNN SDK. Developers can deploy pre-optimized models from the Sixfab Model Zoo or convert custom ONNX models into the DXNN format for execution on the accelerator. Both Python and C++ development environments are supported.

Supported workloads include:

• Object detection
• Image classification
• Instance segmentation
• Video analytics
• Smart surveillance
• Industrial vision systems
• Robotics perception applications

The PCIe-based architecture also provides significantly lower latency and higher throughput compared to many USB-connected AI accelerators commonly used with Raspberry Pi systems.

Comparison with Raspberry Pi AI HAT+

In terms of AI acceleration, the Sixfab AI HAT+ enters a market currently served by the Raspberry Pi AI HAT+ family based on Hailo processors. While both solutions focus on edge AI inference and computer vision workloads, the Sixfab board leverages the DEEPX DX-M1 architecture and software ecosystem.

The 25 TOPS DX-M1M variant offers similar AI performance levels to many dedicated edge AI accelerators while maintaining power consumption in the 2W to 5W range. However, the platform is primarily designed for computer vision applications and is not optimized for running large language models (LLMs) locally due to memory limitations. Future DEEPX products are expected to expand support for additional AI workloads.

Edge AI Expansion Board Also Announced

Alongside the AI HAT+, Sixfab has also revealed an upcoming Edge AI Expansion Board for Raspberry Pi 5. The expansion board combines DEEPX AI acceleration with NVMe SSD storage support and LTE/5G connectivity, targeting industrial and remote edge deployments that require storage, networking, and AI processing in a single platform.

The Sixfab AI HAT+ for Raspberry Pi 5 is available in two configurations:

• DX-M1ML (13 TOPS) variant
• DX-M1M (25 TOPS) variant

The board is available through Sixfab's online store and distribution channels. Developers interested in higher levels of integration may also want to watch for the upcoming Edge AI Expansion Board, which combines AI acceleration, storage, and cellular connectivity into a single Raspberry Pi 5 add-on platform.

With up to 25 TOPS of AI performance, PCIe connectivity, and support for the Raspberry Pi ecosystem, the Sixfab AI HAT+ provides an interesting option for developers looking to deploy computer vision and edge AI workloads directly on Raspberry Pi 5 hardware without relying on cloud-based inference.