How To Run DeepSeek R1 Locally (With ONE COMMAND!)

deep seek R1 is taken to World by storm deep seek R1 is taken to World by storm but if you use it online you're sending but if you use it online you're sending but if you use it online you're sending all of your data to the Chinese which all of your data to the Chinese which all of your data to the Chinese which some people find problematic in this some people find problematic in this some people find problematic in this video I'm going to show you how to run video I'm going to show you how to run video I'm going to show you how to run deep seek locally with one command let's deep seek locally with one command let's deep seek locally with one command let's jump in I prepared this repository for jump in I prepared this repository for jump in I prepared this repository for you on GitHub the link will be down you on GitHub the link will be down you on GitHub the link will be down below and if you open the link you click below and if you open the link you click below and if you open the link you click on code here click the link and then you on code here click the link and then you on code here click the link and then you just open up the terminal and you write just open up the terminal and you write just open up the terminal and you write get clone and you clone the get clone and you clone the get clone and you clone the URL and when it is cloned you type CD URL and when it is cloned you type CD URL and when it is cloned you type CD deep SE local and then you're in in the deep SE local and then you're in in the deep SE local and then you're in in the repo and then the next command that you repo and then the next command that you repo and then the next command that you need to run as I said here in the read need to run as I said here in the read need to run as I said here in the read me is Docker compos profile CPU up so me is Docker compos profile CPU up so me is Docker compos profile CPU up so this is now going to pull in the this is now going to pull in the this is now going to pull in the required images and it will take a a required images and it will take a a required images and it will take a a little while for it to complete but what little while for it to complete but what little while for it to complete but what you need to watch out for is this phrase you need to watch out for is this phrase you need to watch out for is this phrase pulling manifest over here so pulling pulling manifest over here so pulling pulling manifest over here so pulling manifest it's not showing a uh like a manifest it's not showing a uh like a manifest it's not showing a uh like a timer or something like that but pulling timer or something like that but pulling timer or something like that but pulling manifest means that it's now pulling in manifest means that it's now pulling in manifest means that it's now pulling in the model so make sure that you see that the model so make sure that you see that the model so make sure that you see that uh it takes a while there's no progress uh it takes a while there's no progress uh it takes a while there's no progress bar or unfortunately but just give it a bar or unfortunately but just give it a bar or unfortunately but just give it a few minutes okay so when you see this few minutes okay so when you see this few minutes okay so when you see this when you see the AMA pool container when you see the AMA pool container when you see the AMA pool container finished with exit Code Zero then you finished with exit Code Zero then you finished with exit Code Zero then you are ready to test the are ready to test the are ready to test the application to access the application application to access the application application to access the application you open a new browser Tab and you type you open a new browser Tab and you type you open a new browser Tab and you type Local Host colon Local Host colon Local Host colon 8080 and now here we have our open web 8080 and now here we have our open web 8080 and now here we have our open web UI you click get started and then you UI you click get started and then you UI you click get started and then you can just write test test test test.com can just write test test test test.com can just write test test test test.com and password test this is just a local

and password test this is just a local and password test this is just a local account create an admin account and now account create an admin account and now account create an admin account and now we are in our UI let's go we see that we are in our UI let's go we see that we are in our UI let's go we see that our deep seek R1 is already loaded and our deep seek R1 is already loaded and our deep seek R1 is already loaded and let's ask it a question hello what is a let's ask it a question hello what is a let's ask it a question hello what is a banana just asking a simple uh simple banana just asking a simple uh simple banana just asking a simple uh simple question hello a banana is yellow and question hello a banana is yellow and question hello a banana is yellow and here we go this is deep seek running here we go this is deep seek running here we go this is deep seek running locally on my computer on a CPU no GPU locally on my computer on a CPU no GPU locally on my computer on a CPU no GPU required awesome so the interface that required awesome so the interface that required awesome so the interface that we are using here to connect to our we are using here to connect to our we are using here to connect to our local deep seek is called open web UI local deep seek is called open web UI local deep seek is called open web UI there are several features that you can there are several features that you can there are several features that you can use you can um you can also adjust the use you can um you can also adjust the use you can um you can also adjust the controls here how many tokens you want controls here how many tokens you want controls here how many tokens you want wanted to use Etc and there's also voice wanted to use Etc and there's also voice wanted to use Etc and there's also voice features which won't work because I'm features which won't work because I'm features which won't work because I'm using my microphone right now but uh in using my microphone right now but uh in using my microphone right now but uh in at at home for you it should work just at at home for you it should work just at at home for you it should work just fine out of the box so now let's test a fine out of the box so now let's test a fine out of the box so now let's test a couple of interesting prompts right so couple of interesting prompts right so couple of interesting prompts right so who was Julius Caesar and why was he so who was Julius Caesar and why was he so who was Julius Caesar and why was he so special I open a new chat special I open a new chat special I open a new chat paste that in and now we it's it's going paste that in and now we it's it's going paste that in and now we it's it's going to process it and if I open up btop then to process it and if I open up btop then to process it and if I open up btop then here we see that it's now using 50% of here we see that it's now using 50% of here we see that it's now using 50% of my CPU over here 55% so the container is my CPU over here 55% so the container is my CPU over here 55% so the container is kept at using half of all of my kept at using half of all of my kept at using half of all of my processor course and it's now processor course and it's now processor course and it's now formulating the response as I'm waiting formulating the response as I'm waiting formulating the response as I'm waiting and checking out the the UI here and the and checking out the the UI here and the and checking out the the UI here and the cool thing about deep seek is that it's cool thing about deep seek is that it's cool thing about deep seek is that it's a reasoning model so you can actually a reasoning model so you can actually a reasoning model so you can actually view it thinking now so it's saying okay view it thinking now so it's saying okay view it thinking now so it's saying okay so I need to figure out who Julius so I need to figure out who Julius so I need to figure out who Julius Caesar was and why he's special first Caesar was and why he's special first Caesar was and why he's special first off Julius Caesar was a famous

off Julius Caesar was a famous off Julius Caesar was a famous historical figure so you can actually historical figure so you can actually historical figure so you can actually see how deep sake is going to arrive at see how deep sake is going to arrive at see how deep sake is going to arrive at its um at its conclusion which is super its um at its conclusion which is super its um at its conclusion which is super interesting to actually observe how it's interesting to actually observe how it's interesting to actually observe how it's uh reaching that so now it has finished uh reaching that so now it has finished uh reaching that so now it has finished the thinking block and now it's actually the thinking block and now it's actually the thinking block and now it's actually going to formulate the response here and going to formulate the response here and going to formulate the response here and it's working pretty well and I mean I it's working pretty well and I mean I it's working pretty well and I mean I have a pretty beefy CPU I'm running a have a pretty beefy CPU I'm running a have a pretty beefy CPU I'm running a AMD ryzen 9950 50x but as you see the AMD ryzen 9950 50x but as you see the AMD ryzen 9950 50x but as you see the CPU setup is it's working but it is CPU setup is it's working but it is CPU setup is it's working but it is quite slow so I'm not even going to ask quite slow so I'm not even going to ask quite slow so I'm not even going to ask it to um to do the second question it to um to do the second question it to um to do the second question because the second question I had because the second question I had because the second question I had prepared was write me a 500w SE about prepared was write me a 500w SE about prepared was write me a 500w SE about juliia Caesar well this is going to take juliia Caesar well this is going to take juliia Caesar well this is going to take a long time as because we see a response a long time as because we see a response a long time as because we see a response like this is already taking a while so like this is already taking a while so like this is already taking a while so let's check out how we can run this let's check out how we can run this let's check out how we can run this using our GPU but before I show you that using our GPU but before I show you that using our GPU but before I show you that I just wanted to mention that if you run I just wanted to mention that if you run I just wanted to mention that if you run into any problem when setting this up if into any problem when setting this up if into any problem when setting this up if it somehow doesn't work for you or if it somehow doesn't work for you or if it somehow doesn't work for you or if you have any questions make sure to jump you have any questions make sure to jump you have any questions make sure to jump into my cubecraft community the link into my cubecraft community the link into my cubecraft community the link will be down below here you can ask me will be down below here you can ask me will be down below here you can ask me questions me in the community questions questions me in the community questions questions me in the community questions directly and I will answer them you can directly and I will answer them you can directly and I will answer them you can also Join one of the Q&A calls that I also Join one of the Q&A calls that I also Join one of the Q&A calls that I host every week and you get 40 hours of host every week and you get 40 hours of host every week and you get 40 hours of worth of courses here on containers worth of courses here on containers worth of courses here on containers kubernetes Linux and there are also some kubernetes Linux and there are also some kubernetes Linux and there are also some AI courses coming up so make sure to AI courses coming up so make sure to AI courses coming up so make sure to check it out as I said I am running an check it out as I said I am running an check it out as I said I am running an AMD ryzen 950x but I also have a AMD ryzen 950x but I also have a AMD ryzen 950x but I also have a gigabyte 1490 OC GPU here so let's check gigabyte 1490 OC GPU here so let's check gigabyte 1490 OC GPU here so let's check out how we can run this using the GPU so

out how we can run this using the GPU so out how we can run this using the GPU so before you can use your GPU you need to before you can use your GPU you need to before you can use your GPU you need to upgrade your container runtime and in upgrade your container runtime and in upgrade your container runtime and in the read me in the repo here there is a the read me in the repo here there is a the read me in the repo here there is a link included you will need to do a link included you will need to do a link included you will need to do a little bit of extra magic in order to little bit of extra magic in order to little bit of extra magic in order to make use of your Nvidia GPU but when you make use of your Nvidia GPU but when you make use of your Nvidia GPU but when you have done that then all you need to do have done that then all you need to do have done that then all you need to do is to close your container apps here and is to close your container apps here and is to close your container apps here and then it's a good practice to clean those then it's a good practice to clean those then it's a good practice to clean those up up up so down so you run Docker compos profile so down so you run Docker compos profile so down so you run Docker compos profile CPU down and then you remove the volumes CPU down and then you remove the volumes CPU down and then you remove the volumes and the orphant so now it's just and the orphant so now it's just and the orphant so now it's just completely removing everything and I can completely removing everything and I can completely removing everything and I can now uh run it with the profile GPU so now uh run it with the profile GPU so now uh run it with the profile GPU so let me show you how to do that I you run let me show you how to do that I you run let me show you how to do that I you run this command profile Nvidia GPU up and this command profile Nvidia GPU up and this command profile Nvidia GPU up and when I do that it's going as we see here when I do that it's going as we see here when I do that it's going as we see here in the logs it has detected my Nvidia in the logs it has detected my Nvidia in the logs it has detected my Nvidia GeForce rtx4 GeForce rtx4 GeForce rtx4 90 and now it's going to be pulling in 90 and now it's going to be pulling in 90 and now it's going to be pulling in the Manifest again as you see down here the Manifest again as you see down here the Manifest again as you see down here pulling manifest so we have to wait a pulling manifest so we have to wait a pulling manifest so we have to wait a couple of couple of couple of minutes all right so now my setup is minutes all right so now my setup is minutes all right so now my setup is rebuilt using the GPU and when I now rebuilt using the GPU and when I now rebuilt using the GPU and when I now open my open web UI again then we'll see open my open web UI again then we'll see open my open web UI again then we'll see that it has reset so because I remove my that it has reset so because I remove my that it has reset so because I remove my volumes so I can just enter a new user volumes so I can just enter a new user volumes so I can just enter a new user again here we go so now we're back into again here we go so now we're back into again here we go so now we're back into open web UI and we see our deep seek is open web UI and we see our deep seek is open web UI and we see our deep seek is loaded again but if I now answer the loaded again but if I now answer the loaded again but if I now answer the question who was Julius

question who was Julius question who was Julius Caesar then we see that we get an almost Caesar then we see that we get an almost Caesar then we see that we get an almost instant response because now it's using instant response because now it's using instant response because now it's using my GPU to um formulate the response so my GPU to um formulate the response so my GPU to um formulate the response so let's check out a bit of a more involved let's check out a bit of a more involved let's check out a bit of a more involved prompt like write me a 500w essay about prompt like write me a 500w essay about prompt like write me a 500w essay about Julia Caesar Etc uh explaining the Julia Caesar Etc uh explaining the Julia Caesar Etc uh explaining the origin of the phrase Crossing the origin of the phrase Crossing the origin of the phrase Crossing the Rubicon and as I'm doing that I will Rubicon and as I'm doing that I will Rubicon and as I'm doing that I will open up my btop again and now we see open up my btop again and now we see open up my btop again and now we see that my GPU is now being utilized for that my GPU is now being utilized for that my GPU is now being utilized for 92% do you see that over there if I go 92% do you see that over there if I go 92% do you see that over there if I go to the let me see the GPU tab here so to the let me see the GPU tab here so to the let me see the GPU tab here so now we'll see that is uh oh it's already now we'll see that is uh oh it's already now we'll see that is uh oh it's already finished the response actually so here finished the response actually so here finished the response actually so here is my 500w essay um which looks pretty is my 500w essay um which looks pretty is my 500w essay um which looks pretty good the phrase Crossing the Rubicon good the phrase Crossing the Rubicon good the phrase Crossing the Rubicon emerged as a metaphor for Caesar's Boldt emerged as a metaphor for Caesar's Boldt emerged as a metaphor for Caesar's Boldt actions during these campaigns cool so actions during these campaigns cool so actions during these campaigns cool so this is just running locally on my this is just running locally on my this is just running locally on my machine and it's producing great output machine and it's producing great output machine and it's producing great output and this is just a 7 billion parameter and this is just a 7 billion parameter and this is just a 7 billion parameter model so let's try a couple of other model so let's try a couple of other model so let's try a couple of other prompts and it's working so fast that I prompts and it's working so fast that I prompts and it's working so fast that I have to like switch to btop really quick have to like switch to btop really quick have to like switch to btop really quick but let's do a another one about the but let's do a another one about the but let's do a another one about the French Revolution and I click new chat French Revolution and I click new chat French Revolution and I click new chat over here paste that in and I'm just over here paste that in and I'm just over here paste that in and I'm just going to switch to btop here we go so going to switch to btop here we go so going to switch to btop here we go so now we see that my GPU is um not being now we see that my GPU is um not being now we see that my GPU is um not being used at all or like 5% currently because used at all or like 5% currently because used at all or like 5% currently because I'm just rendering these uh applications I'm just rendering these uh applications I'm just rendering these uh applications and if I now send the message and switch and if I now send the message and switch and if I now send the message and switch to btop now we see that my GPU kicks in

to btop now we see that my GPU kicks in to btop now we see that my GPU kicks in like this it's pulling 333 watts and like this it's pulling 333 watts and like this it's pulling 333 watts and it's going at 91% now and when this goes it's going at 91% now and when this goes it's going at 91% now and when this goes down then we know that our response has down then we know that our response has down then we know that our response has been finished so it's taking a bit more been finished so it's taking a bit more been finished so it's taking a bit more time because I asked it a bit deeper time because I asked it a bit deeper time because I asked it a bit deeper question to analyze the French question to analyze the French question to analyze the French Revolution and now we see the GPU is not Revolution and now we see the GPU is not Revolution and now we see the GPU is not being used anymore so I know that I'm being used anymore so I know that I'm being used anymore so I know that I'm finished here so what's cool is that I finished here so what's cool is that I finished here so what's cool is that I can again I can check out the thinking can again I can check out the thinking can again I can check out the thinking process okay so I need to examine the process okay so I need to examine the process okay so I need to examine the significance of this French Revolution significance of this French Revolution significance of this French Revolution now thinking about Etc it just goes now thinking about Etc it just goes now thinking about Etc it just goes pretty deep into it and now here is our pretty deep into it and now here is our pretty deep into it and now here is our final response about the causes of the final response about the causes of the final response about the causes of the Revolution Etc now another great thing Revolution Etc now another great thing Revolution Etc now another great thing about the open web UI is that it's about the open web UI is that it's about the open web UI is that it's actually really well built and it has actually really well built and it has actually really well built and it has many of the features that you would many of the features that you would many of the features that you would expect of for example when you use Claud expect of for example when you use Claud expect of for example when you use Claud then it will preview art you can preview then it will preview art you can preview then it will preview art you can preview code artifacts and things like that so code artifacts and things like that so code artifacts and things like that so let's give it another uh prompt um to let's give it another uh prompt um to let's give it another uh prompt um to create a landing page for a web shop create a landing page for a web shop create a landing page for a web shop that specializes in bespoke Split that specializes in bespoke Split that specializes in bespoke Split keyboards use a classic design that keyboards use a classic design that keyboards use a classic design that reminds us of 1950s American magazines reminds us of 1950s American magazines reminds us of 1950s American magazines Etc that just came up with some Etc that just came up with some Etc that just came up with some something so I'm just entering this and something so I'm just entering this and something so I'm just entering this and I was creating a landing page in I was creating a landing page in I was creating a landing page in HTML here is the code block that it's HTML here is the code block that it's HTML here is the code block that it's putting out and you see how quick this putting out and you see how quick this putting out and you see how quick this is when it's running on my is when it's running on my is when it's running on my GPU and here we ALS also see the GPU and here we ALS also see the GPU and here we ALS also see the artifact already being produced and if I artifact already being produced and if I artifact already being produced and if I open this up then yeah here we have our open this up then yeah here we have our open this up then yeah here we have our landing page is it the best no is it

landing page is it the best no is it landing page is it the best no is it very pretty no but it's working and I'm very pretty no but it's working and I'm very pretty no but it's working and I'm not sending my data to chining as to not sending my data to chining as to not sending my data to chining as to China as I'm creating this so this is China as I'm creating this so this is China as I'm creating this so this is really cool now I'm now only using the 7 really cool now I'm now only using the 7 really cool now I'm now only using the 7 billion parameter model so this I I billion parameter model so this I I billion parameter model so this I I chose this because that will run on most chose this because that will run on most chose this because that will run on most CPUs modern CPUs now let's check out how CPUs modern CPUs now let's check out how CPUs modern CPUs now let's check out how we can change that to running a larger we can change that to running a larger we can change that to running a larger model so in order to do model so in order to do model so in order to do that let's go to ama.com because the that let's go to ama.com because the that let's go to ama.com because the setup that I'm using with the one setup that I'm using with the one setup that I'm using with the one command is under underwater is using command is under underwater is using command is under underwater is using olama as a pre-built container so we can olama as a pre-built container so we can olama as a pre-built container so we can run any model that is um listed here on run any model that is um listed here on run any model that is um listed here on the olama website now deep seek R1 is on the olama website now deep seek R1 is on the olama website now deep seek R1 is on the top and if we go to let's say the top and if we go to let's say the top and if we go to let's say 32 billion parameters 32 billion parameters 32 billion parameters then I can just oh let's then I can just oh let's then I can just oh let's try so now I was running the 7 billion try so now I was running the 7 billion try so now I was running the 7 billion now let's try the 14 billion and here it now let's try the 14 billion and here it now let's try the 14 billion and here it gives us the command AMA Run Deep seek gives us the command AMA Run Deep seek gives us the command AMA Run Deep seek but then all we need to do is we copy but then all we need to do is we copy but then all we need to do is we copy these last couple of words and then these last couple of words and then these last couple of words and then going back to our git going back to our git going back to our git repository then we open our Docker repository then we open our Docker repository then we open our Docker compost. yaml and then we look to this compost. yaml and then we look to this compost. yaml and then we look to this Docker this olama pool in it container Docker this olama pool in it container Docker this olama pool in it container so here this AMA pool here is where the so here this AMA pool here is where the so here this AMA pool here is where the model is specified so we just Chang this model is specified so we just Chang this model is specified so we just Chang this from Deep seek 7B to deep seek 14b and from Deep seek 7B to deep seek 14b and from Deep seek 7B to deep seek 14b and now when I rerun this I just do compose

now when I rerun this I just do compose now when I rerun this I just do compose up Docker compose profile GPU Nvidia up up Docker compose profile GPU Nvidia up up Docker compose profile GPU Nvidia up it's going it's going it's going to up that again it's going to start the to up that again it's going to start the to up that again it's going to start the containers again it's now going to pull containers again it's now going to pull containers again it's now going to pull in a new manifest which is going to take in a new manifest which is going to take in a new manifest which is going to take a while again because as we saw on AMA a while again because as we saw on AMA a while again because as we saw on AMA it was like 14 it was like 14 it was like 14 GB okay so our new finished pulling and GB okay so our new finished pulling and GB okay so our new finished pulling and if I now open my open web UI again then if I now open my open web UI again then if I now open my open web UI again then we'll see that in my models here I can we'll see that in my models here I can we'll see that in my models here I can now open up my new model so you just add now open up my new model so you just add now open up my new model so you just add it to the docker compos file and then it to the docker compos file and then it to the docker compos file and then you can switch to it in deep seek in you can switch to it in deep seek in you can switch to it in deep seek in open web UI I'm going to open up a new open web UI I'm going to open up a new open web UI I'm going to open up a new chat and let's try that landing page chat and let's try that landing page chat and let's try that landing page prompt again and again now we're running prompt again and again now we're running prompt again and again now we're running a a model that has the double amount of a a model that has the double amount of a a model that has the double amount of parameters so maybe we'll get a bit of a parameters so maybe we'll get a bit of a parameters so maybe we'll get a bit of a better response this time with seeds better response this time with seeds better response this time with seeds thinking thinking thinking again again again and I am wondering what the result is and I am wondering what the result is and I am wondering what the result is going to be it's now outputting a code going to be it's now outputting a code going to be it's now outputting a code in a nice code block I really like this in a nice code block I really like this in a nice code block I really like this open webui it's super powerful and actually look at that this powerful and actually look at that this actually looks a lot better like The actually looks a lot better like The actually looks a lot better like The Artisan keyboard here it has a bit more Artisan keyboard here it has a bit more Artisan keyboard here it has a bit more contrast it has some prices here so you contrast it has some prices here so you contrast it has some prices here so you can really see that it somehow has put can really see that it somehow has put can really see that it somehow has put in a lot a bit more thought into this in a lot a bit more thought into this in a lot a bit more thought into this right has a bit of a classic font going right has a bit of a classic font going right has a bit of a classic font going on here it's actually looking pretty on here it's actually looking pretty on here it's actually looking pretty good and this is just running locally on

good and this is just running locally on good and this is just running locally on my machine and again let's try that my machine and again let's try that my machine and again let's try that French revolution essay again new chat French revolution essay again new chat French revolution essay again new chat running deep seek all right so now running deep seek all right so now running deep seek all right so now you've seen how easy it is to run these you've seen how easy it is to run these you've seen how easy it is to run these models locally uh I just want to explain models locally uh I just want to explain models locally uh I just want to explain a little bit how this actually works a little bit how this actually works a little bit how this actually works under the hood because you might be under the hood because you might be under the hood because you might be running this on a Mac and you might be running this on a Mac and you might be running this on a Mac and you might be getting different results so I'm running getting different results so I'm running getting different results so I'm running this on a Linux machine I'm running Arch this on a Linux machine I'm running Arch this on a Linux machine I'm running Arch Linux by the way and when I run so my Linux by the way and when I run so my Linux by the way and when I run so my Linux has my kernel and when I run Linux has my kernel and when I run Linux has my kernel and when I run Docker containers these are able to Docker containers these are able to Docker containers these are able to interact directly with my kernel so interact directly with my kernel so interact directly with my kernel so there is no virtualization layer there is no virtualization layer there is no virtualization layer happening between here now if you're happening between here now if you're happening between here now if you're running on a Mac OS system then if you running on a Mac OS system then if you running on a Mac OS system then if you install Docker uh Docker desktop and I install Docker uh Docker desktop and I install Docker uh Docker desktop and I actually recommend that you use Rancher actually recommend that you use Rancher actually recommend that you use Rancher desktop actually if you don't if you desktop actually if you don't if you desktop actually if you don't if you don't have Docker installed don't have Docker installed don't have Docker installed yet or if you're looking for a solution yet or if you're looking for a solution yet or if you're looking for a solution then I actually recommend that you check then I actually recommend that you check then I actually recommend that you check out Rancher desktop because this is out Rancher desktop because this is out Rancher desktop because this is fully open source works on all platforms fully open source works on all platforms fully open source works on all platforms and um it's works a lot better but then and um it's works a lot better but then and um it's works a lot better but then then you would have this what happens is then you would have this what happens is then you would have this what happens is that you have a Rancher that you have a Rancher that you have a Rancher desktop um layer here so this Rancher desktop um layer here so this Rancher desktop um layer here so this Rancher desktop creates a virtual machine on desktop creates a virtual machine on desktop creates a virtual machine on your Mac MacBook and then the containers your Mac MacBook and then the containers your Mac MacBook and then the containers are created in the virtual machine so are created in the virtual machine so are created in the virtual machine so this is your MacBook this is the kernel this is your MacBook this is the kernel this is your MacBook this is the kernel and then a new virtual machine is and then a new virtual machine is and then a new virtual machine is created on top of that which then has

created on top of that which then has created on top of that which then has you can specify it to have like 16 gabt you can specify it to have like 16 gabt you can specify it to have like 16 gabt of RAM and a few virtual CPUs now this of RAM and a few virtual CPUs now this of RAM and a few virtual CPUs now this desktop this Rancher desktop does not desktop this Rancher desktop does not desktop this Rancher desktop does not know about your GPU so you can't run know about your GPU so you can't run know about your GPU so you can't run these GPU these GPU these GPU accelerated workloads using that kind of accelerated workloads using that kind of accelerated workloads using that kind of setup if you're running on Linux then setup if you're running on Linux then setup if you're running on Linux then you remove this layer here and you can you remove this layer here and you can you remove this layer here and you can actually just talk to your Hardware actually just talk to your Hardware actually just talk to your Hardware directly and therefore also your GPU now directly and therefore also your GPU now directly and therefore also your GPU now there are ways of passing these through there are ways of passing these through there are ways of passing these through but that's beyond the scope of this but that's beyond the scope of this but that's beyond the scope of this tutorial so just so you know that if tutorial so just so you know that if tutorial so just so you know that if you're using this setup on a Mac or you're using this setup on a Mac or you're using this setup on a Mac or Windows machine and you're using like Windows machine and you're using like Windows machine and you're using like Docker desktop or rencher desktop to run Docker desktop or rencher desktop to run Docker desktop or rencher desktop to run your containers then you might be seeing your containers then you might be seeing your containers then you might be seeing degraded performance when using this so degraded performance when using this so degraded performance when using this so let what is all the Deep seek hype about let what is all the Deep seek hype about let what is all the Deep seek hype about so I've been following this for a bit so I've been following this for a bit so I've been following this for a bit now and there are a couple of things now and there are a couple of things now and there are a couple of things that are making deep seek very special that are making deep seek very special that are making deep seek very special currently from what I have gathered so currently from what I have gathered so currently from what I have gathered so they're claiming that they did that they they're claiming that they did that they they're claiming that they did that they trained this model with only $7 trained this model with only $7 trained this model with only $7 million million million and this is $6 million and that's why and this is $6 million and that's why and this is $6 million and that's why the stock prices are now falling the the stock prices are now falling the the stock prices are now falling the Nvidia prices are dropping open AI Nvidia prices are dropping open AI Nvidia prices are dropping open AI prices are Dr dropping because these prices are Dr dropping because these prices are Dr dropping because these American companies are spending spending American companies are spending spending American companies are spending spending billions of dollars and there's this billions of dollars and there's this billions of dollars and there's this Chinese startup that's claiming to do Chinese startup that's claiming to do Chinese startup that's claiming to do this with $6 million so what are we this with $6 million so what are we this with $6 million so what are we spending all this bill all these spending all this bill all these spending all this bill all these billions on in the US right that's the billions on in the US right that's the billions on in the US right that's the question now so that is what why deep question now so that is what why deep question now so that is what why deep seek is so special and why it's taking

seek is so special and why it's taking seek is so special and why it's taking the World by storm now it's going to be the World by storm now it's going to be the World by storm now it's going to be a sign of the future that we're going to a sign of the future that we're going to a sign of the future that we're going to be seeing more of these models coming be seeing more of these models coming be seeing more of these models coming out which take less and less resources out which take less and less resources out which take less and less resources to train and what we're going to see is to train and what we're going to see is to train and what we're going to see is we're going to see some more specialized we're going to see some more specialized we're going to see some more specialized models and that's why I'm also exploring models and that's why I'm also exploring models and that's why I'm also exploring this and wanting to learn how to run this and wanting to learn how to run this and wanting to learn how to run these things these things these things locally and the but the the thing is locally and the but the the thing is locally and the but the the thing is that deeps was trained with such a low that deeps was trained with such a low that deeps was trained with such a low budget but it budget but it budget but it actually uh it matches the performance actually uh it matches the performance actually uh it matches the performance of o1 for example so in my school of o1 for example so in my school of o1 for example so in my school Community someone recently uh shared Community someone recently uh shared Community someone recently uh shared this AI stats website AI artificial this AI stats website AI artificial this AI stats website AI artificial analysis and here we see that here we analysis and here we see that here we analysis and here we see that here we have the latest 01 model by open AI that have the latest 01 model by open AI that have the latest 01 model by open AI that is now scoring 90 on the quality index is now scoring 90 on the quality index is now scoring 90 on the quality index but deep seek is scoring 89 as you see but deep seek is scoring 89 as you see but deep seek is scoring 89 as you see here so it's almost matching the here so it's almost matching the here so it's almost matching the performance of 01 but you can run this performance of 01 but you can run this performance of 01 but you can run this locally on your machine like we have locally on your machine like we have locally on your machine like we have learned just now so as you see we are learned just now so as you see we are learned just now so as you see we are constantly sharing good interesting constantly sharing good interesting constantly sharing good interesting resources and um learning this AI stuff resources and um learning this AI stuff resources and um learning this AI stuff together in my devops community so make together in my devops community so make together in my devops community so make sure to jump into the community if you sure to jump into the community if you sure to jump into the community if you want to learn more about running AI want to learn more about running AI want to learn more about running AI locally and I'll see you in the next one