# 5 Better Ways to Use DeepSeek R1 (Without the DeepSeek Website)

## Метаданные

- **Канал:** Skill Leap AI
- **YouTube:** https://www.youtube.com/watch?v=E2Maj1oouiE
- **Дата:** 27.02.2025
- **Длительность:** 11:50
- **Просмотры:** 10,140

## Описание

💡 Exclusive Monica AI Offer: Get 25% OFF the Unlimited Annual Plan if you sign up within 24 hours, or use my link for 10% OFF → https://bit.ly/4ieBfvz with coupon code SLAI10.

DeepSeek R1 is one of the best AI reasoning models, and there are multiple ways to access it beyond the official website or apps. In this video, I’ll show you five different methods to use DeepSeek R1.

1️⃣ Local Install (Two Ways)
Running DeepSeek R1 locally gives you full control over how you use it. Here are two great options:
Ollama – Runs DeepSeek R1 entirely on your hardware. Easy to set up and fully self-contained.
https://ollama.com
Here is the dedicated DeepSeek R1 Ollama install video: https://youtu.be/5kFV20LatL8

LMStudio – A user-friendly interface for running DeepSeek R1 locally.
https://lmstudio.ai/

2️⃣ Perplexity
Perplexity integrates DeepSeek R1 with search for a powerful AI experience.
https://www.perplexity.ai

Originally available only on the Pro plan
Now free for everyone!

3️⃣ Monica AI
Monica is a multi-model AI assistant with access to multiple AI models in one place.
https://bit.ly/4ieBfvz Use couponcode SLAI10 for 10% off. 
✅ Supports AI models like ChatGPT, DeepSeek, Claude, Llama, Gemini, Grok, and more.
✅ Includes image and video AI tools like DALL·E, Stable Diffusion, Runway, and Pixverse.
✅ Browser extension for:
✔ Summarizing YouTube videos
✔ Email replies in Gmail
✔ Translating web pages
✔ And more!

4️⃣ Groq 
https://groq.com/ provides direct access to the DeepSeek-R1-Distill-Llama-70B model right from their homepage.

5️⃣ Nvidia NIMs
Nvidia offers the fastest way to use DeepSeek R1. If speed is a priority, check it out here. ⚡
https://www.nvidia.com/en-us/ai/

▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
MORE FROM SKILL LEAP:
💡 Join the fastest-growing AI education platform & Instantly access 20+ top courses in AI:
👉 Start with a free trial: https://bit.ly/skill-leap
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

## Содержание

### [0:00](https://www.youtube.com/watch?v=E2Maj1oouiE) Segment 1 (00:00 - 05:00)

I want to show you five different ways for using deep seek R1 without actually using the Deep seek website and some of these are going to give you new functionality and I think you're going to like these more than the website now the first thing I wanted to show you is actually how to run deep seek R1 locally on your computer this is going to install directly on your computer and you could run it privately you could chat with it and it's not going to have any downtime sometimes the website is down this is not going to be down because it's installed on your own computer I'm going to show you two different options LM studio is one of the easier options and there is a website called ama. com and here you could also run different models like deep seek R1 you could download these and install them locally now deep seek R1 actually has these smaller versions of it these are distilled from the big version which is 671 billion parameter it's a massive file pretty much no local computer is going to be able to run this one but you can actually run one of these smaller models and for most use cases the reasoning is actually pretty good so I was able to run the 32b and the 70b now I'm using an M3 MacBook Pro but if you have an older computer you may want to try with the 7B version of it here and then this will kind of walk you through on how to install it this one is a little bit more technical so I'm going to show you the other option in this video but I did make this dedicated video on exactly how to install it so I'll link this in the description if you want to go this route the end result of it looks exactly kind of like chat GPT or the Deep seek website here but the really interesting part about this option here this solama option is you could choose different models and have a chat with multiple models at the same time so I could choose deep card 1 press the plus sign and choose another model that I've also installed that is also open source on my computer so I could run for example llama 3. 2 I could choose that as an option and then if I ask a question over here and send out my prompts it's going to run them side by side here so this way I could compare a reasoning model that's actually going through a thinking process versus just a regular model on top of that I could press the plus sign and upload any type of files I want to interact with over here and this has ton of more advanced options too inside of settings keeps your chat history so if you use any of these chat bots on any website this is going to look very familiar but again it's going to be installed locally and privately on your computer now this other one LM studio is actually far easier to use and install but it's a little B more limited so I still prefer the AMA option but this one is just an app you just download it from Mac windows or Linux now this is what it's going to look like over here and right over here you could start a new chat and then choose the model right on top so you could have deeps car1 for example and other models to that you could add and the way you do that is you go to the Discover section right here and all these models that are available for install are going to be here on the left side there's a couple different versions of deep seek R1 here that you could download and Quinn 2. 5 is actually another really good large language model that is open source 54 this one is from Microsoft that's really good and llama 3. 3 this is from meta that's also really good and then you could install any one of them and then easily compare your answers and switch between these open- Source models too now if you choose R1 here it's going to go through the thinking process it's going to show you how many seconds it's thinking and then it's going to kind of put the answer right over here in this format and give you the final answer on the bottom and you could copy the answer directly ly from this chat down here now this also has this rack function which is when you could upload your own documents so you could upload five documents at a time and these are the different file formats and you could go ahead and interact with those documents as well right inside of LM Studio okay the second option I wanted to show you takes place inside of perplexity AI so perplexity actually has deeps R1 and they have it available in the free account I have the paid Pro account for $20 a month which gives me access to Pro search but this reason with R1 is available with free users so you could choose this one and this version of it is hosted in the US perplexity actually since it's open source model downloaded this and is hosting it here and this combines search with reasoning and specifically the R1 reasoning which is really good so let's ask it a question here and I'll kind of show you how it works so it reasons with R1 and then on top of that it's going to break down your question into smaller questions with each part is going to also search the web so it combines multiple different things in the same prompt you could see it's going through 36 different web sources on top of his own knowledge base on top of the R1 reasoning and he usually gives you incredible answers clot 3. 7 Sonet is better suited for Enterprise environments requiring hybrid reasoning and long context analysis while deep seek R1 dominates the cost sensitive technical applications and open-source customizations now lexity with R1 here I

### [5:00](https://www.youtube.com/watch?v=E2Maj1oouiE&t=300s) Segment 2 (05:00 - 10:00)

think is well suited if you require research they have some other functions with the Pro Plan like deep research too that gives you a much more in-depth research and obviously Pro search gives you three times more sources here if you have that paid upgrade you could always use o03 mini which is the reasoning model from open AI too kind of works like this R1 model so you have a lot of different option but this is really intended for search based tasks so I wouldn't use R1 here if you don't need search at all but this is an option if you want to use it with search one of my favorite ways to do searches right now the third way you could use deep seek R1 here is through another platform called Monica and Monica is an all-in-one platform so it gives you access to all the top AI model so I specifically partnered with them for this video because I wanted to show you some new ways to use this because this could follow you around the web to you don't have to use that on this website I want to show you a couple different things so all the different models you don't have to sign up for all the different apps individually so you can get chat GPT here claw 3. 7 sign this literally came out less than 24 hours ago and they already added it over here obviously you have deep seek R1 so everything is available here you also get deep seek V3 but on top of that you also get access to image generation models too so lot of the top image generation models are going to be available for you to choose from over here you also get video generation models too you could see they have the top video generation models as well so Runway gen 3 one of my favorites cling you could choose those from here this one is actually really good too so you choose that you give it a prompt and you could generate 5sec clips for that now this Chrome extension is actually my preferred way for using this because if I click on it I have that same kind of Monica layout over here that I showed you on the website but it also pulled in the video over here and I could interact the video I could chat with it over here based on the model I have picked over here I could just click on it and I'll get a quick page summary this is just pulling in everything from that video transcription kind of breaking down the key points so if I don't have time to watch this 14-minute video well this is telling me exactly what it covers over here and then I could ask any follow-up questions down here so interacting with YouTube videos with that Chrome extension really powerful and the same thing also works with any web page so you could again click this over here it pulls in the web page over here and then you could interact with that web page now this time I asked for the key points over here and you could choose different models over here you don't have to always choose DC car1 obviously you could choose any of these models by default it's set to this Monica one but look at all these different models that are available every time you ask a question over here in this dropdown and anytime you do a follow-up you could choose your model as well and it also has access to the web too so you could combine models that don't have web access like Claude and then add web access to them this way too so I'll put a link in the description to Monica if you want to try this out again I recommend using a Chrome extension so it kind of follows you around and gives you access to all the models in one bundle thanks for Monica for sponsoring this video Let's jump into number four now this next one is called grock with a Q so there's a Gro with a K that's a whole different thing grock with a Q gives you access to open source models like R1 so you could choose R1 over here I'm going to choose the 70b model over here and then you could go ahead and ask it for any type of reasoning tasks and I'm just going to show you the speed because it beats every single platform out there when it comes to response speed I'll give it a coding task here and I'll just send this out over here you can see it's already thinking and look how fast it's going through that process right now I can't even keep up with scrolling as fast as this is writing it and it's done it literally took seconds here and if you want to see the speed over here this is going to show you the speed right on top and if you're not very technical this is extremely fast right here 27 tokens per second is the actual speed and the whole output took six seconds this probably 5 to 10 times faster than some of the other chat interfaces where you can use deep seek R1 so if you're looking for a speed definitely grock is the place to go for that now at number five this is from Nvidia I haven't mentioned this in other videos but it's a really cool place to go this is called Nvidia Nim okay and these are kind of AI agents that you could build but let me press try it now because this is an entire platform so deep SEC car 1 is right over here you could use it but they have so many different popular models you could choose from so if I press view now you could see all kinds of models some of them you probably never heard of either like this one I actually don't know what this one is but any one of these you could try over here and right now it's completely free to use this platform so I'm going to go back to the model page over here and I'll just choose deep SEC car1 and I'll show you some other things they have that are pretty interesting this is also

### [10:00](https://www.youtube.com/watch?v=E2Maj1oouiE&t=600s) Segment 3 (10:00 - 11:00)

by the way extremely fast this is running Nvidia gpus obviously the top player in the world of AI Computing and you could see this one was almost instant this one might have a wait time though because again it is completely free so I have come across times where I had to wait in line especially if I was generating code but again a really interesting platform but the one reason why I wanted to add it in this video is right here you have parameters so if you are more advanced you could get deeps car1 to change its default behavior that you have in other apps so for example you could change is temperature is if you go to zero it tries to follow the prompt exactly as close as possible and the more towards one you go the more creative it tries to get so you usually try to find a balance in the middle somewhere over here now you have other things too that are also worth looking into but temperature is probably one of my favorite things to tweak to change this response without doing much to The Prompt now while we're here I'll show you some other things they have so they have these kind of AI agents this one turns a PDF into a podcast similar to what notebook LM does this one is a multimodal PDF extractor so you could view all these different models they have these kind of blueprints over here they have Vision language models here that you could explore from the left side over here and in the world of biology they have different ones you could explore in the world in simulations they have all kinds of different workflows and blueprints so I think this is going to be worth exploring but at the very minimum you could always just use deep seek R1 here it's going to be one of the speediest options as long as a lot of people aren't using it at the same time now I also compared deep seek R1 against other top reasoning models so I'll link that video here if you want to see how it's stacked up thanks for watching I'll see you next time

---
*Источник: https://ekstraktznaniy.ru/video/12521*