Raspberry Pi AI HAT+ 2 Review: The brains and the brawn

Raspberry Pi’s first product of 2026 is an update of the 2024 AI HAT+, but this newer version, another collaboration with Hailo, now sees the Hailo 10H AI chip running the show, along with 8GB of onboard RAM. The new AI HAT+ 2 takes the strain of AI workloads away from the Raspberry Pi 5’s Arm CPU, but this all comes at a price of $130. With your Raspberry Pi already costing much more than the original $35 — of course, the spec has vastly improved over the years — you could already be hitting the $200 mark for just a Pi and AI HAT+ 2. Does the performance warrant the price? There's only one way to find out!

Raspberry Pi AI HAT+ 2 Specifications

Unboxing and Setup

Image 1 of 7

The retail box follows the same design language as the many other Raspberry Pi product boxes that I have opened. At a glance, you’d be forgiven for thinking that this was the same Raspberry Pi AI HAT+ as released previously, and opening the box doesn’t help as the boards are very similar. The new AI HAT+ 2 requires the included heatsink. Yes, this heatsink is for the HAT, not the Raspberry Pi 5. Your Pi 5 will also need cooling, and the official Raspberry Pi and Argon low-profile coolers will fit under the HAT. The included plastic standoffs and GPIO header extension work, but the GPIO connection is a little too loose for my liking. The resulting GPIO connections, using DuPoint style connectors, also feel a little too loose.

Connecting the board to the Raspberry Pi 5 is simple. Just unlock the PCIe connection on the Pi 5, push in the ribbon cable, lock it down, and then secure the board to the standoffs and GPIO. There is a cut-out for connecting the official Raspberry Pi Camera and an official display. Connect up your keyboard, mouse, HDMI, Ethernet, and finally power, then boot to the Raspberry Pi desktop, remembering of course to enable PCIe Gen 3 via “raspi-config.” We’re using the latest Debian “Trixie” based image and have a custom installation process as our review unit predates the official software repositories. The end-user software experience will be streamlined for release.

What Models Can the Raspberry Pi AI HAT+ 2 Hailo 10H Run?

Using the provided installation instructions, we ran hailo-ollama and then queried the available and compatible models for the Hailo 10H powering the kit.

In our pre-release software, the models are loaded using hailo-ollama via a carefully crafted curl command. Just change the “model” to one of the five available.

curl --silent http://localhost:8000/api/pull \
     -H 'Content-Type: application/json' \
     -d '{ "model": "qwen2:1.5b", "stream" : true }'

The 8GB of onboard DDR4X RAM means that larger models will generally work better as the Raspberry Pi’s own RAM is untouched. So models up to 8GB should load without incident, even on Raspberry Pi 5s with less than 8GB. This opens up cheaper AI projects, technically.

You still need to pay $130 for the AI HAT+ 2, but a $50 1GB, $55 2GB, or $77 4GB Raspberry Pi 5 is now a viable AI platform, negating the need to buy a $105 8GB or the frankly frightening $160 for a 16GB Raspberry Pi 5.

Comparing the Raspberry Pi AI HAT+ 2 With the AI HAT+

Image 1 of 5

So why the new board? That is down to Large Language Models (LLM), an AI that is trained on huge amounts of text data and is used to understand, process, and respond to human language. The AI HAT+ 2 is mainly aimed at LLMs, whereas the older AI HAT+ is for image-based AI projects. The AI HAT+ 2 demo code supplied by Raspberry Pi leans heavily on creating our own local LLM using qwen2:1.5b but you can also use DeepSeek or Qwen models that are distilled via DeepSeek.

The onboard 8GB of RAM and powerful AI processing chip take the strain off the Raspberry Pi 5’s CPU and RAM. We can also use that power for image processing. If you’ve not got the original AI HAT+, then having good image processing and a viable LLM platform makes the $130 price tag easier to swallow.

The two boards may look similar, but they don’t work in the same way. The AI HAT+ was all about image-based AI processing, and the 26 TOPS of the onboard Hailo 8L (13 Tops for the cheaper Hailo8 model) is very similar in performance to the AI HAT+ 2. Using the rpicam-hello suite of examples, we tested compatibility with the AI HAT+ 2, and we are pleased to say that object identification and pose detection worked as expected. While we have no metrics to show the difference, the performance of the AI HAT+ 2 was smooth, and there were no camera issues. Where there were issues, they were largely software-based. "HailoRT not ready!" shows that Hailo's software is playing catch-up to the hardware, which has happened before. Hailo, Raspberry Pi and the community will ultimately fix this gap, but early adopters need to be aware.

Comparing AI HAT+ 2 and Raspberry Pi 5 CPU Performance

Because we cannot directly test the AI HAT+ 2 and the AI HAT+, we need to fallback to testing the AI HAT+ 2 versus the Arm CPU of the Raspberry Pi 5. For this test, we will use the qwen2:1.5b language model and pose some questions relating to general knowledge, then specific programming tasks using Python. Installation of qwen2:1.5b was made via ollama.

The first question is “What is Tom’s Hardware?”

AI HAT+ 2 Response

“Tom's Hardware is a hardware retailer with stores in the United States and Canada. The company sells various types of tools, equipment, and supplies for home repairs or construction projects.”

ARM CPU Response

“Tom's Guide is a popular technology website founded in 2004. It was originally focused on computer hardware and reviews, but has since expanded its coverage to include consumer electronics, software, and other topics related to technology. The site is known for publishing detailed product reviews that are valuable resources for consumers looking to make informed purchasing decisions.”

Writing Code with an LLM

The next request, “Write a Python script to display an image on the screen. The image file is supplied by asking the user for input and they will specify the absolute path to the image.”

The AI HAT+ 2 did a decent job of writing quick and concise code, but it was doomed to failure as it never called an application to open the image, rather it read the image and then closed the open image file.

AI HAT+ 2 Response

import os
def display_image(image_path):
    try:
        # Open the image file in read mode.
        img = open(image_path, 'r')
        img.seek(0)
        img.read()
        img.close()
    except IOError:
        print('Image file not found:', image_path)
image_path = input('Enter the path to image file: ')
display_image(image_path)

The Arm CPU response looked decent, but on closer inspection, it was full of errors around using the imported tkinter module and calling “Image.open” when there is no imported Image module. This vibe coding experience produced a long-winded response, but ultimately it was way off the mark and not something that we would rely on as a coding co-pilot or a sounding board for a project.

Arm CPU Response

import os
from tkinter import *
# Ask user to enter path to image file
path = input("Enter absolute path to image file: ")
# Load image from disk (assuming this is a JPEG)
img = Image.open(path)
# Resize image if needed
width = 1000
height = img.height / img.width * width
img = img.resize((width, height))
# Create new canvas for the resized image
canvas = Canvas(width=500, height=300, bg="white")
canvas.pack()
# Draw image on canvas and place it in center
image_data = ImageData(img)
canvas.create_image(250, 150, anchor=CENTER, data=image_data)
root.mainloop()

In our tests, the AI HAT+ 2 was faster than the Raspberry Pi 5’s Arm CPU, but more importantly, it ran the code without hogging the CPU. This is great for those who want to integrate AI into a GPIO-based project, like robotics. That said, the model produced inaccurate results. In the case of the coding exercise, the code would appear valid to a layman, but it was completely incorrect. If you are looking to run an LLM on a Pi, then try the Hailo-compatible models and see which one meets your needs. But be warned, the knowledge on which these models have been trained is now outdated, and from our limited testing time, we only saw incorrect responses.

Who is the Raspberry Pi AI HAT+ 2 for?

Obviously, someone who wants to use AI on a Raspberry Pi, but what type of AI? Offloading the workload from the Arm CPU to the Hailo 10H frees up the CPU for other tasks, such as running a chat server, controlling a robot, reacting to sensors, etc. So those of us who like to build smart GPIO projects will have a field day with the AI HAT+ 2.

If you are just interested in image or vision-based AI projects, the older AI HAT+, Raspberry Pi AI Camera, or the original M.2 AI Kit are all cheaper viable options. If you already have any of these products, stick with them, as right now the AI HAT+ 2 is more money for little to no performance boost. If you haven’t got any AI HATs or want to dabble with LLMs, then the AI HAT+ 2 is a viable, if currently flawed, option. Personally, we would run LLMs on the Raspberry Pi 5s Arm CPU until we have the knowledge and use case to warrant purchasing the AI HAT+ 2.

Bottom Line

AI is the buzzword that isn’t going away, and Raspberry Pi’s adoption of AI into its product range is an interesting, if polarizing, decision. The AI HAT+ 2 continues the progression of more powerful AI platforms, and for the right type of make,r it will be a considered choice. One day. Right now, this is a solution looking for a problem, and we're sure that the bugs will be worked out, but early adopters will be left wanting more.

For many who just want to dabble with AI on their Raspberry Pi 5, then they can either use smaller models that your RAM can accommodate, or use an online service. For computer vision and image inference projects, you will get similar performance and a cheaper product with the older AI HAT+ or the Raspberry Pi AI Camera. The AI camera is a cheap entry point for learners. For those who want a local LLM in a compact and power-efficient package, the Raspberry Pi AI HAT+ 2 is something that you should research, after learning the skills and developing the project that it can support. It will also give the software time to mature and to make sure that your wallet is ready.