Raspberry Pi has announced the launch of its new AI HAT+ 2 accessory, a complementary board for the Raspberry Pi 5 that enhances its capabilities for generative artificial intelligence systems at a modest price. This innovative board features the Hailo-10H neural accelerator and 8 GB of dedicated RAM, allowing large language models (LLMs) to be run directly on the device without needing an internet connection. Below, we delve into all the essential details.
Raspberry AI HAT+ 2 Technical Specifications
| Feature | Raspberry AI HAT+ 2 |
|---|---|
| Processor | Hailo-10H AI (40 TOPS inference performance at INT4) |
| Memory | 8GB RAM |
| Workload | Computer vision inference and generative AI models |
| Software Integration | Full integration with Raspberry Pi camera stack |
| Compatibility | Compatible with Raspberry Pi 5 |
| Facility | Includes 16 mm stackable header, spacers, and screws (optional heatsink) |
| Price | $130 |
The price may vary. We earn commission from these links.
A Plate for Local Models
The AI HAT+ 2 represents a significant upgrade from its predecessor, which was launched just over a year ago. The original AI HAT+ primarily focused on computer vision tasks such as object detection and scene segmentation, while the new model aims to meet the growing demand for running language models locally and offline.
The board boasts an impressive 40 TOPS inference performance (at INT4) due to the Hailo-10H chip. The addition of 8 GB of LPDDR4X RAM is a noteworthy innovation, allowing the Raspberry Pi 5’s processor and RAM to focus on other tasks while the neural accelerator handles AI processing.
Beyond Paper
The Hailo 10H chip is limited to a power consumption of just 3W, enhancing energy efficiency compared to the Raspberry Pi 5’s native chip. However, this limitation results in lower raw performance when juxtaposed with the Raspberry Pi 5. Furthermore, although the added 8GB of RAM is substantial, the 16GB version of the Raspberry Pi 5 remains a more powerful option, given the RAM requirements of many medium-sized language models, which often exceed 10 GB.
Target Audience
The AI HAT+ 2 retains the computer vision capabilities of its predecessor and can run vision and language models simultaneously. It is an ideal fit for applications that demand both vision processing and language model inference, particularly in industrial automation systems or IoT devices. Additionally, the board serves as an excellent development kit for companies wishing to integrate the Hailo-10H chip into their own products.
For enthusiasts seeking to run local language models, purchasing a Raspberry Pi 5 with 16GB of RAM may remain the most practical choice.
Price and Availability
The Raspberry Pi AI HAT+ 2 is available for $130, although price variations may occur depending on local distributors. Furthermore, resources, tutorials, and documentation are accessible in the Hailo GitHub repository and the Developer Zone.
Raspberry Pi AI HAT+ 2, 8 GB RAM (add-on board to Raspberry Pi 5)
The price may vary. We earn commission from these links.
Images | Raspberry Pi
In Xataka | Someone has used a Raspberry Pi worth less than 10 euros for a great project: creating their own original Macintosh.