Raspberry Pi AI HAT+ 2 Brings Gen AI to Raspberry Pi 5
A new version of the Raspberry Pi AI HAT+ has been released with enough grunt to run popular generative AI (GenAI) models, including Qwen and DeepSeek.
The first generation Raspberry Pi AI HAT+ add-ons were designed for on-device acceleration of vision-based neural network models, but wasn’t powerful enough to run more generalised large-language models (LLMs) – something the new AI HAT+ 2 solves.
Described as the company’s “first AI product designed to fill the generative AI gap”, the Raspberry Pi AI HAT+ 2 packs in a Hailo-10H NPU with 40 TOPS (INT4) of inferencing performance and is paired with 8GB of dedicated on-board RAM.
“Performing all AI processing locally and without a network connection, the AI HAT+ 2 operates reliably and with low latency, maintaining the privacy, security, and cost-efficiency of cloud-free AI computing that we introduced with the original AI HAT+”, say the company.
Vision-based models can be run on the new board too, with AI HAT+ 2’s computer vision performance described as “broadly equivalent to that of its 26-TOPS predecessor”. It retains tight integration with the native Pi camera software stack.
Below are the LLMs a Raspberry Pi 5 with the AI HAT+ 2 will be able to run at launch:
LLMParametersDeepSeek-R1-Distill1.5 billionLlama3.21 billionQwen2.5-Coder 1.5 billionQwen2.5-Instruct 1.5 billionQwen21.5 billion
Newer and in some cases larger models will become available soon, and the Pi community is almost certain to enjoy experimenting with creating and tuning ‘quantised’ variants of open source models to wring every bit of performance out for specific tasks.
In a video demo shared by Raspberry Pi, the Qwen2 model predicts cogent seeming answers — that is what an LLM does; it’s not plucking answers out of thing air, but statistically guesstimating an applicable answer from training data — to a couple of basic questions:
Other demos shows the coding tasks, translation and, marginally more interesting as a fop, a Vision Language Model (LVM) describing what it sees from a camera stream:
Performance (never mind accuracy) will not rival that of cloud-based AI giants which boast up to 2 trillion parameters in their models and can harness huge amounts of compute.
But smaller, more nimble models have benefits: focus (it’s often said creativity is borne of constraint); and processing data on-device, rather than someone else’s cloud, offers privacy and security benefits.
How much does it cost?
The Raspberry Pi AI HAT+ 2 costs $130/£125 and is available to buy from approved Pi resellers from January 15, 2026.
Don’t need the power on offer here? The first-gen AI HAT+ models remain available and are cheaper, priced around ~£68 for 13 TOPs and ~£105 for 26 TOPS respectively. Just keep in mind that those models aren’t designed for running GenAI tasks.
