Running DeepSeek AI on a Raspberry Pi

In this video, we quickly explore how you can run the distilled versions of DeepSeek R1 on a Raspberry Pi.
Tutorial: pimylifeup.com...
Ollama Tutorial: pimylifeup.com...
Below is a list of some of the equipment we used when running DeepSeek on our Raspberry Pi. The faster your Pi is and the more RAM it has the better experience you are going to have with running this AI model locally.
Please note that we will get a small commission if you buy from any of these Amazon links.
Raspberry Pi: amzn.to/3WUT8Ij
Micro SD Card: amzn.to/3US0ZUz
Ethernet Cable: amzn.to/3K8BcT7
While the Raspberry Pi is definitely not capable of running the full blown DeepSeek R1 model, it is actually able to handle some of the smaller distilled variants.
These distilled variants are part of what has made DeepSeek R1 such a popular AI to run locally. Simply put, a distilled AI is one that takes the full-blown model, then uses a bit of smarts and other AI to cut it down to what should still be a very viable AI, but significantly easier to run.
Some of the distilled variants of DeepSeek have been cut down enough that they can even run on a Raspberry Pi. The smallest is actually close to real-time speed but suffers greatly in accuracy and has a tendency to hallucinate answers. However, the Pi can still some of the larger but still heavily distilled variants that generate some very usable results.

[@Maximum443]

I have also run deepseek r1 on my rpi 5, but with a different method, using docker and ollanma running a webui. This allowed me to run the full model, although it was quite slow and cpu temps went crazy.

[@syntartica]

I love your website full of content that we can use for our raspberry. Keep it up. Thanks for the videos too. ☀💪

[@BadKarma714]

Love this guy‘s videos, am I the first comment? Lol How can I host my own website on a Raspberry Pi web server?

[@akaxa000]

Can u explain me what do you mean by hosting own website? You will use raspberry pi to run a website? Am i wrong?

[@Catchgate]

Just what I needed, thank you, I was thinking about getting a Pi AI Hat+ - have you any experience with this, or thoughts as to how useful this would be against the Pi without the Hat? Would be curious on your thoughts!

[@Pimylifeup]

The Pi AI Hat+ won't give you any boost in performance with these LLMs (Large language models). The processor/accelerator they use (Hailo-8) isn't really designed for processing these models and is also not supported within the Ollama software that we are using.

[@marci1270]

This is not Deepseek R1. Its just a distiled model ....

[@MNGermann]

4:06 my 1st test question is usually “what does the fox says?”