Cómo usar DeepSeek Mejor que el 99% de las personas (GRATIS) — Transcript
Full transcript
- 0:00It’s not always the one with the best
- 0:01cards who wins, but the one who learns
- 0:03how to play the hand they have.
- 0:05Even when it seemed they had the worst
- 0:08hand, DeepSeek found a way to be more
- 0:10efficient and optimized in the
- 0:11artificial intelligence market. Today,
- 0:14they don't just have the best AI models
- 0:16in terms of price-to-performance, but
- 0:19also a complete ecosystem capable of
- 0:21anything.
- 0:25To prove it, we’ll cover everything
- 0:27about DeepSeek, from zero to expert,
- 0:29with everything you need to know, like
- 0:32their models, apps, key functions, and
- 0:34running it locally or uncensored.
- 0:37The best part is that everything
- 0:39DeepSeek developed is open, meaning we
- 0:41can start from cheap to free, private,
- 0:44and limitless.
- 0:45Sound good?
- 0:47Well, let’s start at the beginning.
- 0:51Let’s properly place DeepSeek. This is a
- 0:54Chinese company founded in 2023 and
- 0:56based in Hangzhou. Behind it is Liang
- 0:59Wenfeng, an entrepreneur and engineer
- 1:01who worked in artificial intelligence.
- 1:04This project was born linked to High
- 1:06Flyer, a quantitative investment fund
- 1:08that used AI to trade in financial
- 1:10markets.
- 1:11Its goal was very clear: to research and
- 1:14develop artificial general intelligence
- 1:16to compete against the big labs.
- 1:19Their great revolution arrived in
- 1:21January 2025 with DeepSeek R1, which
- 1:23shocked everyone with its reasoning
- 1:25capacity and efficiency, causing the US
- 1:28artificial intelligence market to
- 1:30plummet.
- 1:32They are committed to working with
- 1:34highly optimized, open-source models
- 1:36that can be downloaded, modified, and
- 1:38executed.
- 1:40Looking at their business model, it
- 1:42might seem like everything is free,
- 1:44since we really have a free chat,
- 1:46low-cost APIs, and open-source models
- 1:49you don't have to pay for.
- 1:51But here is the key: even though their
- 1:53models are low-cost, they can support it
- 1:56because they are very efficient, and
- 1:59that open-source code also helps the
- 2:01project improve future features. Let’s
- 2:04start by getting to know their models a
- 2:06bit more in depth.
- 2:07Basically, we have two main ones: the
- 2:09Flash models, which are the fastest,
- 2:12cheapest, multimodal, and best for
- 2:14day-to-day tasks; and the Pro model,
- 2:16which is larger, geared toward complex
- 2:19tasks and deeper reasoning.
- 2:22Although as of the making of this video
- 2:24, the latest update is the 4.1 Flash
- 2:25version, which, besides being cheaper,
- 2:27is also more powerful.
- 2:29Historically, we also find the R1 models
- 2:32, the family I mentioned before, focused
- 2:34on mathematical, logical, and
- 2:36programming reasoning. And from this,
- 2:39distilled models were born, also known
- 2:41as R1 Distils, which are smaller
- 2:44versions designed to run on fewer
- 2:46resources. And from these, many
- 2:48community variations have also emerged,
- 2:51even releasing them to be free of any
- 2:53type of restriction. DeepSeek also
- 2:56develops other types of models like OCR
- 2:58and UL, but we will focus mainly on the
- 3:01AI ones that we can use through its
- 3:03ecosystem.
- 3:05To make use of this model, the easiest
- 3:07way would be to jump to their official
- 3:08cloud platform with their DeepSeek chat
- 3:10.
- 3:11From here, keep in mind that we will be
- 3:13able to use their latest model, and also
- 3:15, no matter where you look, you won't
- 3:16find a paid subscription since they
- 3:18offer it completely for free.
- 3:21Operation is very simple; we simply
- 3:22input any prompt here, as in this case
- 3:25where I'm going to ask for a
- 3:26mini-application to show companies how
- 3:28much each meeting is costing them by
- 3:30calculating it based on the number of
- 3:32attendees, hourly rate, duration, and so
- 3:34on.
- 3:36When I click send, this tool starts
- 3:38working, and a minute later, after all
- 3:40the reasoning we can see here, a little
- 3:42further down we can see that it has
- 3:44developed the entire code for the
- 3:46application, which we can also preview
- 3:48right here.
- 3:50We just have to click Run, and from
- 3:52there we can find this meeting cost
- 3:55calculator, where we can input the
- 3:57number of attendees—for example, I'll
- 4:00put seven attendees—the hourly cost,
- 4:03I'll use 75 as an example, and an
- 4:05approximate duration of an hour or so.
- 4:09It shows us very clearly how the cost
- 4:11would be about 525 €, and it also
- 4:14developed a very simple timer so that
- 4:16when we hit play, we can see in real
- 4:19time the cost that meeting is incurring
- 4:22, something I think could be very
- 4:24interesting for more than one company.
- 4:27And from this platform, the truth is
- 4:29that we have a very capable model via
- 4:31chat with which we can do many things
- 4:33with practically no limits at all.
- 4:36However, if we want to get the most out
- 4:38of DeepSeek, we should start with the
- 4:40next section.
- 4:44In this second section, we are going to
- 4:46talk about one of the main pillars that
- 4:48DeepSeek currently has, which is its
- 4:50application known as DeepSeek Harness.
- 4:53And this harness is climbing the ranks
- 4:55and positioning itself among the top,
- 4:57comparing it with the main tools like
- 5:00Codex, Hermes, Open Code, Cloud Code,
- 5:02and others.
- 5:04And I'm not surprised, because not only
- 5:06has it managed to keep up in just a few
- 5:08months and in record time, but it has
- 5:10also incorporated certain features that
- 5:12none of the previous ones have. To
- 5:14verify this, I went to the official
- 5:16DeepSeek Harness page, and by the way,
- 5:18all the links I’m mentioning throughout
- 5:20this video will be provided in the
- 5:22description below.
- 5:24From here, keep in mind that as of today
- 5:26, we have a very simple way to download
- 5:28its application.
- 5:30We just have to click here, and if for
- 5:31some reason it doesn't show your
- 5:33operating system correctly, if we click
- 5:35this arrow, we can see that it is
- 5:36available for both Mac and Windows.
- 5:39Once downloaded and installed, we will
- 5:42find an interface very similar to any
- 5:44harness like ChatGPT, only this one is
- 5:47completely open source.
- 5:49Before looking at the key points of this
- 5:51desktop application, in case you are
- 5:53wondering, yes, we also have an app for
- 5:56mobile devices.
- 5:57To do this, go to your mobile phone's
- 5:59app store, click on the search bar, and
- 6:01type "DeepSeek."
- 6:03From here, you'll find this official app
- 6:06, which you can install for free.
- 6:09Once opened, after logging in or
- 6:11creating an account, it will confirm
- 6:13this is the official free app, and by
- 6:16clicking "start chat," you can start
- 6:18using DeepSeek right away.
- 6:20But you should keep in mind that the
- 6:23mobile app is more similar to the web
- 6:25version with its chat mode, rather than
- 6:28what we see on the computer where we
- 6:30actually work with agents.
- 6:32If we return to the application, it is
- 6:35true that we have some additional
- 6:36features, such as the fact that by
- 6:38holding down this button, we can dictate
- 6:41the prompt more easily.
- 6:43Now, if I let go, we can see how it
- 6:45responded instantly thanks to the
- 6:47transcription of what I just said.
- 6:50Let's return to the desktop application
- 6:52now, since this is really where we have
- 6:54a ton of possibilities.
- 6:56To discover all of this, let's look at
- 6:58the next block: the key points to keep
- 7:00in mind to get the most out of it.
- 7:02However, I will end this on a relatively
- 7:04negative note, which is the fact that we
- 7:07don't have that free DeepSeek artificial
- 7:09intelligence model linked here, but we
- 7:11will have the possibility to connect it
- 7:14officially via API at a very low cost
- 7:16and with all the agentic possibilities
- 7:18that we will be seeing in the next block
- 7:21. But beyond this, thanks to the fact
- 7:23that it is open source, we will be able
- 7:24to integrate any provider, even those
- 7:26that offer us the ability to use it for
- 7:28free.
- 7:30From here, notice how I actually have a
- 7:31bunch of artificial intelligence models
- 7:33that I can use.
- 7:35Many of them, moreover, if I type the
- 7:37word free, I will be able to use for
- 7:39free, and for that, keep in mind that if
- 7:41it is the first time you download and
- 7:43install this application, it will
- 7:44probably ask you to add some credit
- 7:46during the installation process.
- 7:48This is not necessary, you can skip it
- 7:50since I am now going to tell you all the
- 7:52ways to connect artificial intelligence
- 7:54. To do this, what we have to do is
- 7:56click on the sidebar on the left and
- 7:58from here, if we go to the plus sign and
- 8:00then enter the settings section, we will
- 8:02find an option called models.
- 8:04From here, we will be able to connect
- 8:06any AI provider, whether it is the
- 8:08official one itself or any other tool
- 8:10that offers us artificial intelligence
- 8:12models. These can be in the cloud or
- 8:14even installed locally, as we will be
- 8:16seeing later.
- 8:18To be able to add a provider, we simply
- 8:20have to click on this section here,
- 8:21select some from this list in the
- 8:23providers section, but be careful,
- 8:25because if you don't find a provider,
- 8:26you can also add it in a custom way, as
- 8:28we will see later with local models.
- 8:31To see the official case, I am going to
- 8:33select the DeepSeek one and from here
- 8:35all we have to do is paste our API key.
- 8:38I already have it pasted, but keep in
- 8:40mind that you will be able to do this by
- 8:42jumping to the official DeepSeek
- 8:44platform, which I will also be leaving
- 8:45below in the description, and by adding
- 8:47some credit here, even if it is just
- 8:49about 2 dollars, which actually goes a
- 8:51long way, then by entering the API Key
- 8:53part and clicking on create a new API
- 8:55key, by entering a name and selecting
- 8:57create, we will only have to copy, paste
- 8:59from here and when we click apply we
- 9:01will have the model available.
- 9:03However, the best thing about having
- 9:05this open-source model is the ability to
- 9:07go beyond a specific provider. To do
- 9:10this, notice how I have Nvidia installed
- 9:12here, I even have my own ChatGPT codec
- 9:15and others like OpenRouter.
- 9:18I've done this by following the exact
- 9:20same process, but using these other
- 9:21providers that also offer free APIs.
- 9:24With what we have just seen, we are
- 9:26ready to do a lot with all these
- 9:27artificial intelligence models. In large
- 9:30part also due to the information found
- 9:32on the internet to train their AI models
- 9:34. Now, have you ever wondered how much
- 9:36of your personal information is
- 9:38circulating on the internet without you
- 9:40knowing?
- 9:41To find out, you should know about
- 9:43Incogni, a service that takes care of
- 9:45requesting data deletion for us, because
- 9:48there are companies that collect and
- 9:50sell our personal data, like our phone
- 9:52number, email, or even our address.
- 9:56I don't know if it has happened to you,
- 9:57but I keep getting calls from
- 9:58telemarketers and scammers, but we can
- 9:59avoid this this way.
- 10:02Look, if we simply create an account and
- 10:04authorize it to act on our behalf, from
- 10:06that point on it will connect with more
- 10:08than 420 data brokers requesting its
- 10:10removal.
- 10:11In addition, it repeats requests
- 10:13periodically and we can check the
- 10:15progress at any time.
- 10:17I did it with mine and it showed me
- 10:19everything easily, even with a privacy
- 10:21score, but although it looked good, I
- 10:22discovered a bunch of brokers that were
- 10:24selling my personal information.
- 10:27Only now Incogni is removing it
- 10:29automatically. And hey, thanks to
- 10:31Incogni being sponsors of this video,
- 10:33they have left me a special link that
- 10:35also has a bigger discount for Black
- 10:37Friday, which I will leave for you in
- 10:38the description below, with which by
- 10:40also using the code Ale Javi you will
- 10:42get an even bigger discount. Take
- 10:44advantage because these are the best
- 10:45prices of the year and this is one of
- 10:47the tools that is worth trying. And once
- 10:49we know this, let's move on to the next
- 10:50block.
- 10:54I have just returned to the DeepSeek
- 10:56application to really see those keys
- 10:58that make it different from other
- 11:00platforms. One of the things to keep in
- 11:02mind is that the main thing that
- 11:04differentiates it is this plugins
- 11:06functionality, since it literally allows
- 11:08us to develop any function we want or
- 11:10even integrate one from the community
- 11:13that we like.
- 11:14To see this, I'm going to click on
- 11:16plugins and from here notice how we will
- 11:18have a lot of functionalities, both
- 11:20available and some experimental ones
- 11:22that we could activate or deactivate
- 11:24simply by clicking here.
- 11:26But beyond all these functionalities
- 11:29with sub-agents and so on, we will also
- 11:31be able to integrate the ones we need.
- 11:34Now imagine that you like to use some
- 11:36free tokens offered by some artificial
- 11:38intelligence platforms.
- 11:40For example, Open Code offers some free
- 11:42models directly on its own platform,
- 11:44like the one from DeepSeek itself.
- 11:47But even if we can't do this directly
- 11:49from here, because it has a large
- 11:51community and is open source, we can
- 11:53integrate it easily.
- 11:55In fact, notice how we already have
- 11:56repositories right here that can do
- 11:58exactly that.
- 12:00And to integrate any functionality we've
- 12:03seen in a repository, it's as easy as
- 12:05clicking "add plugin," pasting the
- 12:08GitHub URL here, and simply clicking "
- 12:10install" to start processing it so it
- 12:13later appears here, ready for us to
- 12:15activate.
- 12:18Now, if I start a new session and from
- 12:20the model section, select those from the
- 12:23Open Code provider, notice that if I
- 12:25search for "DeepSeek," beyond the
- 12:28official ones I have via API, I can also
- 12:30find free versions like DeepSeek 4.1
- 12:33Flash.
- 12:35If I click on this "free" one and ask it
- 12:37if it can hear me, as you can see, after
- 12:40about two seconds, it replies "yes"
- 12:42perfectly.
- 12:44The plugins really allow us to add
- 12:46whatever we want, not just free
- 12:47artificial intelligence—and by the way,
- 12:50if you’d like to know all this in detail
- 12:52, this video here talks exactly about
- 12:54all the options DeepSeek gives us to
- 12:56connect for free—but beyond that, if we
- 12:58click on plugins again, we could install
- 13:00anything.
- 13:02Notice how I installed another
- 13:04repository here that allows me to use my
- 13:06own ChatGPT models from my subscription
- 13:09directly on DeepSeek, and along with
- 13:11this, beyond integrating existing
- 13:14repositories, we could create our own
- 13:16plugins.
- 13:18To do that, we can simply click the plus
- 13:20sign, select "create your own plugin,"
- 13:22and by writing what we want it to create
- 13:23here, it will be able to do it.
- 13:26And with this, we can really do things
- 13:29like create a timer to easily see how
- 13:32long each task takes, keep track of all
- 13:34spending in terms of both tokens and
- 13:37cost, completely change the interface,
- 13:40customize it, create interactive 3D
- 13:42applications like this world, or even
- 13:45change the application name, like I did
- 13:47here when I named it "Aleja Bijarnes."
- 13:51I already talked about all these "
- 13:52creator mode" examples in this other
- 13:54video here, so if you want to dive a
- 13:55little deeper, you have that video there
- 13:57, but let's look at other key features
- 13:59of the application.
- 14:01The thing is, focusing on a new chat
- 14:03session, one of the lesser-known aspects
- 14:05is the small features geared toward
- 14:07specific tasks, as well as the types of
- 14:09formats we can preview from here.
- 14:12Let’s start with the basics; keep in
- 14:14mind that through our harness, we can
- 14:16grant access to any folder on our
- 14:18computer, whether for it to review or
- 14:20execute tasks, such as reorganizing
- 14:22everything, moving files, creating new
- 14:25ones, and more. We will also have
- 14:26different modes available right here.
- 14:28By default, we have the standard mode,
- 14:31but there are others like PTC, focused
- 14:33on massive tasks; the minimal mode,
- 14:35where we force users to utilize the
- 14:37terminal to test and verify performance
- 14:40more efficiently; and the creator mode,
- 14:42which I mentioned earlier, allowing us
- 14:44to change the application to our liking
- 14:47and add any features we want.
- 14:50Let’s start with a simple task, so I’ll
- 14:53go back to the standard section, and
- 14:55before sending it, notice we also have
- 14:57this last option that allows read-only
- 14:59access—very useful if we want to give
- 15:01access to a folder with a lot of data
- 15:03that shouldn't be modified, even if we
- 15:05want it to use that information.
- 15:08By clicking on read-only, we gain that
- 15:10control and avoid any modification
- 15:13errors. Another option allows it to
- 15:15execute changes.
- 15:16We also have a third one that grants
- 15:18full access, allowing it to make all
- 15:20changes without constantly asking us for
- 15:23confirmation.
- 15:24And finally, a beta self-review mode so
- 15:26it can iterate on itself before giving
- 15:29us the final result. Knowing this, I’m
- 15:31going to write the following prompt,
- 15:33asking it to create a visual
- 15:34presentation about six artificial
- 15:36intelligence scams and how we can
- 15:38protect ourselves from them.
- 15:40To do this, I will ask it to include
- 15:42eight 16: 9 slides with all those key
- 15:44points and real statistics.
- 15:46I’ll click send, and it will start
- 15:48processing it right away.
- 15:50In this case, we are using the free Open
- 15:53Code model, and about seven minutes
- 15:55later, it would have created a
- 15:57presentation and even a PDF that we can
- 15:59preview here, so let’s click on the
- 16:01presentation first, and see how the
- 16:03presentation appears for us to expand,
- 16:06giving us all the details we requested.
- 16:10Actually, you can skip through the video
- 16:12if you want to see those common AI scams
- 16:14we're encountering, but let's focus on
- 16:17the features.
- 16:18If I close from here, notice that if I
- 16:20go to the PDF section now, it’s also
- 16:22something I’ll be able to preview and
- 16:25even download.
- 16:26And we aren’t just talking about
- 16:28creating applications here; we can
- 16:30actually interact with any type of
- 16:32documentation, application, or folder
- 16:34with control and a very simple way to
- 16:36interact. Returning to the presentation
- 16:39, I want to show you another thing that
- 16:40sets this DeepSeek application apart,
- 16:42which would be the entire traceability
- 16:44and transparency section.
- 16:47Notice that when we go to any chat, like
- 16:49this one here, by clicking on the
- 16:51trajectory, we can see the breakdown of
- 16:54everything that happened in the
- 16:56background, what tools were used, what
- 16:58the assistant was doing, and we can
- 17:00truly expand and detail this entire
- 17:03timeline to see exactly what was going
- 17:05on.
- 17:07Because yes, beyond being able to
- 17:08customize it to our liking, we can also
- 17:10have control over everything that is
- 17:12happening.
- 17:13Additionally, notice that if we go back
- 17:16to the settings section, we also have an
- 17:18AI model usage section where we can see
- 17:21all the costs incurred according to the
- 17:23different providers, like these free
- 17:25models I'm using from Open Code, Codex
- 17:28itself, and ultimately a breakdown of
- 17:30everything happening at any given moment
- 17:33.
- 17:38In this fourth block, we’re going to
- 17:39focus more on local usage, because while
- 17:41what we’ve seen so far is great and
- 17:43gives us a lot of control, what happens
- 17:44when we want greater privacy or to
- 17:46simply use models that are totally
- 17:47unlimited?
- 17:49Well, we can do this too, not only
- 17:51because we can install local AI models
- 17:54in our own DeepSeek application, but
- 17:56because there are also open-source
- 17:59DeepSeek models that we can integrate
- 18:01with their own harness very easily.
- 18:04Before learning about any model or
- 18:06wasting time installing something that
- 18:08might slow down your computer or just
- 18:10not work well, I’m going to recommend a
- 18:12site that will tell you if it will
- 18:14actually work before you do all of that
- 18:16, so you can save all that time. The
- 18:19platform is this one here, which I’ll
- 18:20also be leaving down in the description
- 18:22. And from here, when we enter, the
- 18:24first thing it will do is analyze our
- 18:26hardware.
- 18:27In my case, it says I have an Apple M3
- 18:29Pro with 18 GB of VRAM, with the speed
- 18:32we have and the cores it contains.
- 18:34Knowing all this, it can then recommend
- 18:37which models will run best on my device
- 18:39. Notice how it recommends, for example
- 18:41, small models like Alibaba's Qwen,
- 18:43models like Melun, and some from other
- 18:45companies like Google, Mistral, and
- 18:47others.
- 18:49But here, if we wanted to use a specific
- 18:52model, like DeepSeek, we can go to the
- 18:54search section and, by clicking on the
- 18:57search bar and typing the model we want
- 18:59to install—in our case, DeepSeek—all the
- 19:02available ones we can use will appear.
- 19:06Notice how the ones we had via API, like
- 19:08Flash and Pro, actually require so much
- 19:11hardware that practically no computer,
- 19:13even if it's open source, will be able
- 19:16to run them at home. And that is why, if
- 19:18we want much more power, it might be
- 19:20interesting to rely on the previous API
- 19:22blocks. However, for slightly simpler
- 19:25tasks, we can use smaller models. For
- 19:27this, we have R1 and distilled R1, which
- 19:30have fewer parameters and allow us to
- 19:32run them on our device easily.
- 19:34In my case, I could run more or less up
- 19:36to the 14-billion parameter version, and
- 19:39it shows me how easily my device will
- 19:41handle it. So, once we know this, we are
- 19:44ready to install artificial intelligence
- 19:46models locally. To do this, we have many
- 19:48possibilities. We have platforms like
- 19:51Jan, LM Studio, or the one I like best,
- 19:53Ollama.
- 19:55Being this platform right here—which, by
- 19:57the way, I'm going to go a bit faster
- 19:59since I talked about it in detail in
- 20:01this video here—I'll give you some key
- 20:03points: click to download and install
- 20:05this application, because once you have
- 20:07it, even if it looks similar to DeepSeek
- 20:09, just in a more basic way, what we are
- 20:12really interested in is installing the
- 20:14models locally and then transferring
- 20:16these models to DeepSeek itself.
- 20:18Let's see how to do this. For that, I've
- 20:20returned to Ollama, and we'll start by
- 20:22seeing how we can download an artificial
- 20:23intelligence model. We will do this by
- 20:25clicking on the models section and
- 20:27searching from there for the model we
- 20:28want to install. To do this, I am going
- 20:31to search for DeepSeek and go to a
- 20:33slightly older model, the R1, since it
- 20:35has different variations depending on
- 20:37the number of parameters it has.
- 20:40In my case, since you can use it with up
- 20:42to 14 billion parameters, I will stick
- 20:44with this last one with 8 billion
- 20:46parameters, so I will click on copy
- 20:48model and from a terminal—and pay
- 20:50attention, because even though it may
- 20:52seem a bit technical, it is very simple
- 20:54—just by typing the word "Ollama pull,"
- 20:56we add a space and paste the ID we just
- 20:58copied for the model we want to download
- 21:00; upon pressing enter, it will start
- 21:02downloading.
- 21:04Notice how it directly shows us the
- 21:05download speed and how long it will take
- 21:07; in my case, approximately a minute and
- 21:09a half, but this will depend a lot on
- 21:11your computer and the type of model you
- 21:13are downloading. After that time, it
- 21:15will tell me here that it has been
- 21:16successfully downloaded, and to test it
- 21:18, the best thing would be to return to
- 21:20the Ollama application, where that model
- 21:22should appear.
- 21:23See how I have it here, DeepSeek R1 with
- 21:258 billion parameters, and if I send it
- 21:27any test like "hello, how are you?"
- 21:30And after having reasoned the answer
- 21:31thoroughly, we can see that it has
- 21:33responded well, and we will also be able
- 21:35to use this without any internet
- 21:37connection, as everything will stay
- 21:38between the computer and us.
- 21:41So now we are going to see how we can
- 21:43connect this model to our own DeepSeek
- 21:44application, where we have more advanced
- 21:46features.
- 21:47To do this, notice that if we go to the
- 21:50left sidebar and then to applications,
- 21:52we will find a bunch of options here to
- 21:54connect local models to many artificial
- 21:57intelligence tools. Among them, we have
- 21:59the DeepSeek harness, and to do this,
- 22:01all we have to do is click on this
- 22:04command named "Ollama launch dsh," so
- 22:06here in a new terminal, I will paste
- 22:08that command and press enter. From here
- 22:10, it asks us which model we want to
- 22:12connect to.
- 22:14In fact, here we can find some models
- 22:16that they offer on their own cloud, or
- 22:18others that we can install locally or
- 22:19those we already had before, and if we
- 22:21also wanted to install one directly from
- 22:23here, like Qwen, for example, just hit
- 22:25enter and click that we want to download
- 22:28it; after doing so, our DeepSeek
- 22:29application would reload, only with that
- 22:31model connected.
- 22:33I just canceled it because I want to
- 22:35show you another way, since it's really
- 22:36not the only method. If we go to the
- 22:39more section and enter the settings menu
- 22:41again, remember that we can also add a
- 22:44custom provider for models. From here,
- 22:47we can enter the data to connect to our
- 22:49Ollama.
- 22:51To do this, for the name I'll enter
- 22:53Ollama, for the base URL we'll put this
- 22:55one here, and by clicking on search
- 22:57available models, all the ones we have
- 22:59installed locally will appear, such as
- 23:01Cuen, DeepSeek R1 with 8 billion
- 23:03parameters, and Gemma 4.
- 23:06If we wanted to use all three, we could
- 23:08leave them checked, but if instead we
- 23:10only wanted to use some of them, like
- 23:12I'm going to leave these two here,
- 23:14DeepSeek and Gemma, I can click on
- 23:16update list.
- 23:17They would then appear linked here, and
- 23:20if I click on create provider, I'll have
- 23:22this custom provider. Then, when I click
- 23:25on the models section, beyond all those
- 23:27available via API, if I scroll to the
- 23:30bottom, I can find these two models that
- 23:32we have installed locally. And with this
- 23:35, we not only get free use, but also
- 23:37unlimited and private access.
- 23:43We will finish with the final block,
- 23:44knowing that we can install DeepSeek
- 23:46models locally, which have also been
- 23:48modified by the community and completely
- 23:50unlocked without restrictions.
- 23:53Use it responsibly because it is
- 23:55dangerously powerful, as it can
- 23:56reverse-engineer applications or answer
- 23:58anything we need.
- 24:00Before seeing how we can use these fully
- 24:03unlocked DeepSeek models, I want to show
- 24:05you some use cases so you can see the
- 24:07opportunity this offers. For one, you
- 24:09can use it to research controversial
- 24:11topics, such as historical or political
- 24:14ones, that some models avoid answering.
- 24:16We can also perform reverse engineering
- 24:18to study how our own applications work,
- 24:20or those we are authorized to analyze,
- 24:23and learn to build alternatives. We can
- 24:25use it for research and cybersecurity,
- 24:27to analyze vulnerabilities in our own
- 24:29systems and understand how to protect
- 24:31them. And we can also use it with total
- 24:33privacy for more sensitive matters,
- 24:35because since it runs on our computer,
- 24:37we can consult confidential documents,
- 24:39discuss personal issues, or explore
- 24:41health questions without sending our
- 24:44conversation to external servers or
- 24:46having it tell us it can't answer that.
- 24:48Regarding the unlocked version, keep in
- 24:50mind that although I recommend it
- 24:51locally, it could also be used in the
- 24:53cloud.
- 24:54If we go to tools like Luteos or
- 24:56Alliteration, from there they will
- 24:58really let us run models that are open
- 25:01source and uncensored.
- 25:03Keep in mind that we will be able to use
- 25:05them with private servers, but there
- 25:07will always be some cost, since we are
- 25:09theoretically running them with more
- 25:11security.
- 25:12However, in many repositories we can
- 25:14find, for example on Hugging Face, we
- 25:17can also find models like DeepSeek
- 25:19version 4 Pro based on GPT-3.5, which is
- 25:21completely uncensored.
- 25:23These also don't require much hardware
- 25:25and we can install them in a similar way
- 25:27or as we have seen in previous steps,
- 25:29only instead of Ollama, we have to do it
- 25:32with other providers like Hugging Face.
- 25:34With all this, DeepSeek shows us that it
- 25:36is capable of doing a lot and that, on
- 25:38many occasions, it is not necessary to
- 25:40have the best hand to win, since
- 25:42sometimes moving forward with what we
- 25:43have can change the course entirely.
About this transcript
This page contains the full transcript of Cómo usar DeepSeek Mejor que el 99% de las personas (GRATIS) by Alejavi Rivera, generated from the public captions YouTube serves with the video. The transcript has 4,660 words across 722 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.