alt Hacker NewsRaspberry Pi's New AI Hat Adds 8GB of RAM for Local LLMs
Comments
I think Raspberry lost the magic of the older Pis, they lost that sense of purpose. They basically created a niche with the first Pis, now they're just jumping into segments that others created and are already filled to the brim with perhaps even more qualified competition.
Are they seeing a worthwhile niche for the tinkerers (or businesses?) who want to run local LLMs with middling performance but still need full set of GPIOs in a small package? Maybe. But maybe this is just Raspberry jumping on the bandwagon.
I don't blame them for looking to expand into new segments, the business needs to survive. But these efforts just look a bit aimless to me. I "blame" them for not having another "Raspberry Pi moment".
P.S. I can maybe see Frigate and similar solutions driving the adoption for these, like they boosted Coral TPU sales. Not sure if that's enough of a push to make it successful. The hat just doesn't have any of the unique value proposition that kickstarted the Raspberry wave.
In the UK I've never seen the hailo hats (which are quite old BTW) advertised for LLMs. The presented usecase has been object detection from lots of video cameras in realtime.
They seem very fast and I certainly want to use that kind of thing in my house and garden - spotting when foxes and cats arrive and dig up my compost pit, or if people come over when I'm away to water the plants etc.
[edit: I've just seen the updated version in Pimonori and it does claim usefulness for LLMs but also for VLMs and I suspect this is the best way to use it].
> In practice, it's not as amazing as it sounds.
8GB RAM for AI on a Pi sounds underwhelming even from the headline
A few years ago this product would have just been called an ML Accelerator and marketed as helping accelerate ML workloads like object detection in images
Hitching their wagon to the AI train comes with different expectations, leading to a mixed bag of reviews like this.
Interesting idea. I think the Jetson Orin Nano is a better purchase for this application. The main downside is the RAM is shared so you lose about 1G from the OS overhead.
Is there any usefulness with the small large language models, outside perhaps embeddings and learning?
I fail to see the use-case on a Pi. For learning you can have access to much better hardware for cheaper. Perhaps you can use it as a slow and expensive embedding machine, but why?
A good illustration of how “can run LLM” ≠ “makes sense to run LLM”. A prime example of how numbers in specs don’t translate into real UX.
As an edge computing enthusiast, this feels like a meaningful leap for the Raspberry Pi ecosystem. Having a low-power inference accelerator baked into the platform opens up a lot of practical local AI use cases without dragging in the cloud. It’s still early, but this is the right direction for real edge workloads.
My primary concern with this offering is vendor of the enabling silicon. I think it’s important to consider who you do business with. The industrial uptake of RPi boards has poisoned their mandate and made them forget making a world where all children can discover the magic of computing, not just the special chosen ones.
Excited to see more hardware competition at this level. Models that can run on this amount of RAM are right in the sweet spot of small enough to train on consumer-grade GPUs (e.g. 4090) but big enough to do something interesting (simple audio/video/text processing).
The price point is still a little high for most tasks but I’m sure that will come down.
can't wait to not be able to buy it, and also for it to be more expensive than a mini-computer
I buy a raspberry pi because I need a small workhorse - I understand adding RAM for local LLMs, but it would be like a raspberry pi with a GPU, why do i need it when a normal mini machine will have more ram, more compute capacity and better specs for cheaper?
Hmm. Can this "AI" hardware - or any other "AI" hardware that isn't a GPU - be used for anything other than LLMs?
YOLO for example.
Sounds like some PM just wanted to shove AI marketing where it doesn't make sense.
This AI 8GB HAT is $150, while better performing Pi5 8GB is $100. Makes no sense to buy the AI HAT.
8GB is really low.
That said, perhaps there is a niche for slow LLM inference for non-interactive use.
For example, if you use LLMs to triage your emails in the background, you don't care about latency. You just need the throughput to be high enough to handle the load.
It does seem like this is a marketing gimmick. I doubt it will have worked in terms of sales numbers. "What else does it do?"
It's useless for LLMs and it's actually slower than Hailo 8H for standard vision tasks, so, why ?
I've seen the AI-8850 LLM Acceleration M.2 Module advertised as an alternative RPi accellorator (you need an M.2 hat for it).
That's also limited to 8Gb RAM so again you might be better off with a larger 16Gb Pi and using the CPU but at least the space is heating up.
With a lot of this stuff it seems to come down to how good the software support is. Raspberry Pis generally beat everything else for that.
Are there significant usecases for the really small LLMs right now (<10b distills and such)?
My impression so far was that the resulting models are unusably stupid, but maybe there are some specific tasks where they still perform acceptably?
Any chance you can split layers so that some run on the CPU and some run on this board to let you run bigger models and/or get better performance?
I wonder -- how does this thing compare against the Rubik Pi [1]?
What not use a USB Coral TPU? Seems to do mostly the same stuff and is half the price.
What a pointless product to waste time making.
i wonder how the Hailo 10H compares to Axera AX8850. add on boards seem to be cheaper and its a full SoC that can also draw much more power.
Can someone share a non-theoretical LLM application that demands local inference and pocketable hardware? The internet hype for these is sky high, but the actual results are disappointing.
Gigabytes?? In THIS economy?
What’s the current state of the art in low power wake word and speech to text? Has anyone written a blog post on this?
I was able to run a speech to text on my old Pixel 4 but it’s a bit flaky (the background process loses the audio device occasionally). I just want to take some wake word and then send everything to remote LLM and then get back text that I do TTS on.
Glad Jeff was critical here they need a bit of a wake up call it seems.
Catalog TG 211, 1000 Hz.
At this moment my two questions for these things are:
1. Can I run a local LLM that allows me to control Home Assistant with natural language? Some basic stuff like timers, to do/shopping lists etc would be nice etc.
2. Can I run object/person detection on local video streams?
I want some AI stuff, but I want it local.
Looks like the answer for this one is: Meh. It can do point 2, but it's not the best option.
I wonder if this is the magic hardware for LiquidAI/LFM2.5-Audio-1.5B
Dont need more than 8gb. It'll be enough power. IT can do audio to audio.
> The Pi's built-in CPU trounces the Hailo 10H.
Case closed. And that's extremely slow to begin with, the Pi 5 only gets what, a 32 bit bus? Laughable performance for a purpose built ASIC that costs more than the Pi itself.
> In my testing, Hailo's hailo-rpi5-examples were not yet updated for this new HAT, and even if I specified the Hailo 10H manually, model files would not load
Laughable levels of support too.
As another datapoint, I've recently managed to get the 8L working natively on Ubuntu 24 with ROS, but only after significant shenanigans involving recompiling the kernel module and building their library for python 3.12 that Hailo for some reason does not provide outside 3.11. They only support the Pi OS (like anyone would use that in prod) and even that is very spotty. Like, why would you not target the most popular robotics distro for an AI accelerator? Who else is gonna buy these things exactly?
This looks pretty nice for what it is. However, the RAM is a bit oversized for the vast majority of applications that will run on this, which is giving a misleading impression of what it is useful for.
I once tried to run a segmentation model based on a vision transformer on a PC and that model used somewhere around 1 GB for the parameters and several gigabytes for the KV cache and it was almost entirely compute bound. You couldn't run that type of model on previous AI accelerators because they only supported model sizes in the megabytes range.
"For example, the Hailo 10H is advertised as being used for a Fujitsu demo of automatic shrink detection for a self-checkout."
... why though? CV in software is good enough for this application and we've already been doing it forever (see also: Everseen). Now we're just wasting silicon.
That seems completely and utterly pointless.
A NPU that adds to price but underperforms a rasp cpu?
You get SBC with 32gb ram…
Nevermind the whole minipc ecosystem which will crush this
Are computers even for anything else anymore? Is it just AI until the bubble pops? This is exhausting.
[dead]
[flagged]
I had a couple of Pis that I wanted to use as a Media center, I always had some small issues that created a suboptimal experience. Went for a regular 2nd hand amd64 with a small form factor and never looked back, much better userspace support and for my use case a much smoother experience, no lags no memory swapping and if needed I can just buy a different memory bank or a different component. I have no plans to use a raspberry pi any time soon. I am not sure these days if they really still have a niche to fill and if yes how large this niche is.
Within minutes of each other I can read a bunch of comments about how an 8GB AI hat for RPi would be grand for LLMs, while another contingent points out that my M2 MAX 96GB MacBook is completely useless for LLMs. All I can say is at least the latter is also a great laptop.