The Raspberry Pi 5 has a new add-on board to boost performance


Raspberry Pi has sold the AI HAT+ board for over a year now, giving Pi boards the ability to run lightweight machine learning models for object detection, file tagging, and other tasks. Now, there’s a Raspberry Pi AI HAT+ 2, and it’s powerful enough to run large language models (LLMs).
The new Raspberry Pi AI HAT+ 2 is an add-on board for the Raspberry Pi 5, powered by a Halio-10H chip and 8GB of dedicated RAM. It can deliver 40 TOPS (IN4) of inferencing performance, which is close to the 40-60 TOPS in the Neural Processing Units (NPU) found in newer PC chips from Intel, AMD, and Qualcomm.
You’ll be able to run large language models like Llama 3.2 1B, DeepSeek-R1-Distill 1.5B, Qwen2.5-Coder 1.5B, Qwen2.5-Instruct 1.5B, and Qwen2 1.5B, and Raspberry Pi says additional models with larger parameter sizes are being developed. The 8GB RAM isn’t enough for models like GPT-OSS or regular DeepSeek, but it wouldn’t be a stretch for Gemma 3 1B or other smaller models to be eventually supported.
The board isn’t just for LLMs, though. It works with the same camera software stack as the original AI HAT+, so you can set up vision-based models for object detection, pose estimation, or other machine learning workflows. Raspberry Pi says performance for vision models like Yolo is “broadly equivalent to that of its 26-TOPS predecessor, thanks to the on-board RAM.”
Raspberry Pi said in a blog post, “cloud-based LLMs from OpenAI, Meta, and Anthropic range from 500 billion to 2 trillion parameters; the edge-based LLMs running on the Raspberry Pi AI HAT+ 2, which are sized to fit into the available on-board RAM, typically run at 1–7 billion parameters. Smaller LLMs like these are not designed to match the knowledge set available to the larger models, but rather to operate within a constrained dataset.”
If you’re working on a project that could benefit from machine learning or generative AI models, like a camera that detects which birds are flying past your window or a text-to-speech engine, a Raspberry Pi 5 with the AI HAT+ 2 could be one way to do it. Hopefully, the list of available models grows over time, so the list of capabilities continues to expand.
You can buy the Raspberry Pi AI HAT+ 2 starting today for $130 from authorized stores. That’s slightly more expensive than the 8GB RAM Raspberry Pi on its own, which now costs $95.
Source: Raspberry Pi



