Although the path may sometimes seem complicated, there are countless routes on the map that will lead to the same destination. While artificial intelligence companies like Anthropic and OpenAI seemed to be tightening their limits and reducing free access, Chinese music companies like DeepSek, Zi, and Quen demonstrated a different path thanks to the optimization of their models. Since then, we've seen headlines about models on par with Fable, but free and accessible, giveaways of millions of tokens in the best models, companies correcting their mistakes by lowering their prices by up to 80%, and even platforms that were simply giving away the latest models.
However, there was a catch: when we started using these more intensively, we encountered instability, as some promotions were temporary, others had hidden limitations, or required minimum hardware. Nevertheless, in the shadows, a project called Omovirrouter made a splash to solve this problem, proclaiming itself as the gateway to free and open-source AI. [Music] To find out, we'll discover what OVNIRuter is, its key features, and how to install it so we can do things like connect free models without any authentication, unify several free models from our own API, or even remove limitations from platforms like Cloud Code, among other use cases.
The best part is that this platform is open source under the Meet license, which means we can use it for free, without limits, and privately. Sounds good? Well, let's take a look.
OVNIRuter is a platform that was recently [Music] launched. It's only been around since the beginning of this summer, but it's already generating a lot of buzz, as it's offering more than 1. 5 billion free tokens monthly, unifying a catalog of more than 300 providers, whether via authentication, API, local models, and more.
But what has made this project known isn't so much the headline, but these three key features. First, it features a 12-factor scoring system to prioritize one model over another, taking into account factors like limits, latency, and other aspects that influence the choice of one free model over another. Second, it includes Context Relay.
This refers to the context of the model when it approaches its limit, and Omni Router creates a brief summary to pass on to the next model so it can continue with the task. Finally, there's the entire optimization section, which is crucial, especially when using free models. On Router can optimize up to 95%, with an average of 89% in each AI iteration.
It achieves this through RTK, which filters the output of large tools, and Cavem, which rewrites the promo code, making it shorter and more concise. This is key, as it allows us to optimize potential limitations on free platforms. If you visit their official website, which I'll link to in the description below, you'll see their approach of never abandoning what you're doing, whether it's programming or simply interacting with artificial intelligence.
To that end, they offer us a multitude of free providers and tools. This allows us to switch to a free model, should we reach the limit with our current one, optimizing all costs and starting from $0, since we'll have access to over 90 free providers, 56 of which will always be free. We'll explore this tool step by step and practically, so let's move on to the first example.
In this first section, we'll see how to install this tool so we can start using it. If we return to the page after clicking on "Start for free," you'll notice it tells us that we literally only need to enter three commands. We could open our computer's regular terminal, although we can also use any other ID.
However, I'm going to use the terminal, so I'll paste the first command here. About 5 minutes later, it will tell us that all the packages are installed. So now we can move on to the second command.
First of all, if you get an error, check if you have this Node JS tool installed beforehand. Since it's necessary for installation, I'll also include it in the description. Going back to where we left off, I'm going to enter the second command.
Press Enter, and you can see how it's taken me directly to the website, even though we're actually running it locally. Notice that it will ask for a password, but it shows us what it would be. In this case, it would be "change me," which we can change later.
So, let's copy this, paste it here, and click Continue. Once that's done, we have access to the platform, which, as you can see, is running on our own computer. The first thing we need to do, for security, is change the password.
Notice that we have this configuration section, and if we click on it, some additional settings will appear, but in this case, we'll click on the security settings. When we do this, we'll have to enter the current password, in this case, "change me," and then enter the new password here. Once entered, we click "Update Password," and the password will be successfully changed.
With that, we'll have access to all the tools available on this platform. Notice that if I expand each of these options, we'll see a multitude of functions. We'll highlight the most important ones, but going back to the home screen, we see the key steps we should take, such as creating the API key, connecting several providers to obtain free models that we can rotate through as we reach the limit for each one , and we'll also find a diagram of all the connected providers.
So, one of the most important sections here is to start by connecting these providers that will give us free models. But let's look at the second case. In this second section, we'll begin by connecting different artificial intelligence models through different providers on our own OVN router.
To do this, I've gone back to where we left off, and we need to click on Omniproxy. And once we're here, we need to click on the third option called "Providers. " Once inside, we'll see that we have almost 300 providers in total, none of which we've connected yet.
So, the first thing to do is learn about the different types of providers. I recommend starting with those that don't require authentication. You can do this simply by clicking on "No Authentication.
" As you can see, we have nine different providers that we can connect to without needing an API key, logging in, or anything that links us to a personal account. This is key because if we misuse the service or if anything happens to our account, like forgetting our password or getting banned, it won't affect us, since with this method we aren't assigning our own account. So, from here, let's try connecting different providers without authentication.
One of the best -known is this one from Open Code. So, to see it, let's click on it. And the first thing I recommend before enabling anything is to test which of these models work.
You can see this directly by clicking play. This sends a request for this model, and as you can see, in this case, we have an error. We can do the same thing consecutively with the other models, but it's best to click on the "test all models" section, and clicking it will do this in bulk.
In fact, notice here how they confirm that we have three models available out of six, as this often changes. Sometimes they release free models that later become unavailable. So , this way, we can check very easily.
Having a provider with different working models, such as Xiaomi's Mimo, Nvidia's Nemotron, and Hunjuan's, makes it a very useful provider to use. To do this, we simply click on "Add Account. " Having clicked there and enabled it, if we go back, we can see that we already have a connected provider with three AI models.
We can repeat this process. We simply have to click again on the type of provider we're interested in. In this case, we'll return to those without authentication.
We'll try another one, like this one here. First, I can see that it has models like Yenma 4 with 31 billion parameters. So, let's click again on "Test Everything" to make that call and see what it returns.
And as we can see here, we have an available model with 24 billion parameters. In addition to the models without authentication, which may seem limited, we also have other methodologies, such as the opposite: models that we can access simply by logging into our account. To do this, it would be as easy as clicking on some of these options, such as the models we have in Google Antigravity.
I'm going to click on it. From here, it will show us how to set up the connection. We simply have to click on "add connection," then "continue," and when we log in with our account, it will directly give us some information, in this case, a URL, which we'll have to paste directly here.
So I'm going to paste it, we click " connect," and we can see how it connected to my account. We can also see from here all the available models, which we'll test again, since Yemini has released new models and some may have been removed. In fact, from here we can see how we can connect with the 3.
6 models, some from Cloud, which we also have through Antigravity, and other models. Here we have 15 out of 15 available, since, as you can see, they all have the green checkmark, which is normal, as everything is usually more up-to-date when we don't use that identification method. Another method I'm going to recommend for adding providers, besides those with or without authentication, is those we have via API.
But instead of filtering by this method, which we already have here, I'm going to directly select all those that offer free models. To do this, we'll click here, and as you can see, we'll have more than 120 tools offering free models. Here we'll find the main tools, but if we wanted to find a specific one, such as Open Router, which is one of the best known, we could simply type it here.
And notice how it would appear directly. In this case, it would be a Bappi connection, so to do this, we would click on Open Router, select "Add Connection," and from there we would enter our API key. We can do this directly on the Open Router platform.
Click "Create API Key" in the top right corner, give it a name ( in this case, Omniir Router), click "Create," and copy the API key. Simply paste it here, and click "Check" to confirm it's valid. You could give it a name for organizational purposes.
I'm going to call it "tuto. " Then click "Save. " Once you've done this, you'll see that we've imported over 400 IAS models.
This way, we're enriching many providers with numerous AI models, which we can then rotate through as we reach the free limit for each one. Of all the providers I'm adding, I'm using completely free accounts, which is really where the main advantage of this tool lies. Keep in mind, though, that you could also reduce costs by using paid platforms, but in my case, since it's a relatively new platform, I prefer to keep the paid accounts separate.
Once this is added, notice that if we go back to the providers section, we already see the three providers I added manually. From here, we could continue searching for other providers like this one I've highly recommended called Grock, which also offers free apps not only for text but also for images, transcriptions , and more. Similarly, we could repeat the process by adding a connection.
Here, we 'll see if it's authentication or automatic; simply click on it. In this case, since I've also gone to this platform and logged into a free account, we'll click on API key. Click on create.
I'm going to create one called "tuto," click upload, copy the API key that was created here, and paste it into the OmniRouter section. After giving it a name and saving it, we 'll find that we have another 13 free models. Click save, and with that, we'll have four different providers with several models connected.
Again, we could spend the entire video connecting free AI models, whether for authentication, BAPI, or no authentication at all, but we really get the idea, so I'm going to stick with these four providers. As we just saw, this gives us a lot of freedom, allowing us to connect different providers and use different models from the same place, without depending on a single platform. And we can apply this same concept to the real world when we travel.
That's why I want to introduce you to Sale, an ESN from NordSecurity, the company behind NOR VPN, with mobile data in over 200 destinations without having to find a different SIM for each country. This is as easy as installing the SIM, choosing a plan from the app, and it activates automatically when you arrive at your destination. Then, if you're going to travel to a different destination, you simply add it to the same ESN and you can reuse that SIM for another destination.
You can even use regional or global plans so you can move between different countries while keeping everything centralized. Furthermore, being part of the NorvPN family, we have a host of interesting tools, such as web protection against malicious domains and trackers, an ad blocker that also helps save data, and virtual locations with over 115 options. This is very useful, as many countries have restrictions on the use of different tools, but with this SIM card, we can remove those limitations.
Just as Omni Router gives us the freedom to connect with any free model, Sil will give us the freedom to stay connected from anywhere. And thanks to Seily, the sponsors of this video, they've provided a special link, which I'll include in the description below, so you can check it out for yourself. Now that we know this, let's look at the third use case.
In this third case, we'll cover everything related to providers, as that would warrant its own video. So, in the next case, we'll put it into practice by unlocking a tool like Cloud Code. But first, I want to recommend some interesting providers to consider.
Going back to our application, one I want to recommend is Orc Router. Even though they're not very well-known, they're being very aggressive in getting themselves noticed. In fact, looking at one of their latest tweets, we can see that they're offering models like Deepsic V4 Pro and the Flash version for free.
But in addition to this, we can see that they're adding new models to this free plan. To do this, again, if you click on "add connection," you'll see that it's very similar to Open Router's; it's Vappi. And going to their platform and seeing all the available models, we can indeed find the free Tcen Hunuan models, as well as some from Deepsic and Quen.
So, to access them, I'm going to click on "get API key. " They're also running a promotion where they're giving away. .
. If we start it from GitHub, so I'm going to do it that way, and create our account. Again, I 'm going to go to the APIs section, click on create new key, I'm going to name it "tuto", click create, and copy and paste it from here.
If we check it, we can see that it's valid, so I'm going to save it. And that way, another 190 models have been imported, including free premium models like Tipsic V4, Quen, and Junuan. So we'll close here, and we'll have it added.
Another one I want to recommend is the Nvidia graphics card APIs, since we have the Nvidia NIN platform directly, for which we can also find APIs. We simply have to click on Try Appi. And from there, by creating a free account, we'll be able to see how they offer other interesting models for free, like GLM 5.
2. We also have the free Deepsek V4 and Nemotron 3 models. So, even though it might seem like the models are repeated, that's actually a good thing, since one of these platforms might remove those models from the free version, and if that happens, you'd switch to the other provider.
So, again, I'm going to click on "Generate an API key," copy it here, and then, under "Providers," I'm going to look for Nvidia, which we have right here: Nvidia NIM. We're going to add the API connection. I'm going to rename it to "tuto," and by clicking "save," we've imported another 102 models.
But be aware that this doesn't mean all 100 are free, but only some of them, which we can also pre-configure. Another provider I want to recommend is one for authentication. Specifically, I'd like to recommend Kiro AI, since we'll be able to see their pro plan once we log in directly to their website.
If we click on upgrade, you'll see that the Pro license would cost us per month. But here's a catch: if we click on upgrade, they've had a promotion practically since they launched: the first month is completely free. I already discussed this in detail in this other video, and it's also very important to keep in mind in this video.
When you have a paid subscription, as I mentioned in the video I recommended earlier, the latest models available in the Cloud, such as Set 5, Opus 5, and others, will be unlocked. Finally, another thing we can do is connect all the models we have installed locally. Keep in mind that until now, even though we're running the platform on our own computer, all the processing is actually done in the cloud directly through the API or authentication we've added.
In other words, it runs on the servers of those providers. So, really, regardless of whether our computer is powerful, it's not necessary up to this point. However, if you have a computer that can run models locally, keep in mind that you'll be able to search for providers here, such as Olama, and if you connect it, you could also connect models you have locally.
However, I'm going to leave it at that regarding providers, so let's put it into practice directly in the next example. In this fourth section, we'll connect our OVN router to all the providers with free models directly on one of the most closed platforms, Cloud Code. To do this, I'm going to start by downloading an ID such as Visual Studio, although you could also use others that you already have installed, like Cursor, Google Antigravity, and others.
You'll be able to install this on Mac, Linux, and other devices, so I 'm going to download it directly on my Mac. While it's downloading up here, keep in mind that even though we're assuming you already have Cloud Code installed on your computer, if that's not the case, this video will also show you how to install it, even if you don't have a paid Cloud Code subscription. Once it's open, it will finish loading here.
An empty folder that I've named "tuto". From here, if we click on "create folder" and name it "cloud point", then inside, when we click on "add file", we have to add the word "settings. jonj".
We press Enter, and from here we would paste all this information that Jason, which I'm going to provide with this step-by-step guide, is making it a bit faster in an explanatory document so you can see how to interconnect Omni Router with Cloud Code. Ultimately, you need to be clear on these first three parts, since on the one hand, we're connecting it to the URL where we have our Omni Router installed. We're telling it to connect to this platform and connecting it with a combined model that we'll see shortly.
This one is called Best Free, because if we go back to the Omni Router platform, as we can see here, it clearly shows all the connected models on the home page. From here we can see that all the AI models are active, except for the one highlighted in red, which is showing an error. We can see this in real time, but the truth is, we have many AI models and providers.
Ideally, instead of connecting to a specific one, we should tell Omnir Router to offer us the best free model from all connected providers. To do this, if we type "combo" into the search bar, we'll find this combo option, and from here we can actually create our own. We'll see this shortly, but keep in mind that if we expand this option, we'll see that we already have different pre-configured combinations, such as the best encoding , vision, chat, and other models.
But the one that interests us most is the "auto" for the best free model. By keeping this option selected, it will automatically rotate through the best models from the providers installed on our Omnir Router. That's why we have it configured this way, as in all subsequent models, so that regardless of whether we upgrade to a more powerful or a lighter one, we'll always be using the free options.
We press Ctrl+S to save. Now, if we type the word "cloud" here in the terminal once it's installed, you'll see how this pre-configuration loads directly with the "best free" plan instead of the paid cloud plans. Also, notice that if I type anything from here, like asking for three quick ways to make money with AI, when I click "send," it will start processing it.
With this prioritization of bundles based on all the providers installed on the Omniir Router, it will send us this response as quickly as you're seeing now, because it's actually selecting the best plan for the task. At this point, after about 22 seconds, it would have finished writing everything, giving us advice and the three main ways to make money: AI-generated digital products, automations and chatouts for business, and AI content creation services. And as you can see here, it would be doing this directly in the "best free" section, although you might also be wondering which models it's using.
Well, if we hover over the OVNIR router, we'll be able to see all the traceability of this. Notice that if we go back to the beginning, we can see how it's directly connecting and working with the Nvidia model, since, as we can see, there were a couple of errors with the Grock and Orca providers, and that's why it switched directly to this other provider that's working correctly. All of this happens without us even noticing, and it's really important that it performs the task we ask it to, but also keep in mind that we can go into the most detailed information.
Notice that we have a monitoring section here, and by viewing all the logs, we'll find a breakdown of everything that's happening. For example, here we might find things like. .
. We've been using the Cloud Code platform right now with the best free bundle, and it takes about 20 seconds. Clicking on it will show us the breakdown, revealing how we've been using the Cloud Opus 4.
6 Thinking model with the Antigravity provider. This eliminates the tedious limitations we encountered with each tool. Now, if one fails, it will automatically switch to the next, allowing for more comprehensive use.
While adding another tool or feature to the existing ones might be overwhelming, this simplifies things considerably. Because we have a single platform, regardless of the providers' offerings, we can forget about updates. We'll have this unified tool that selects the best model for each situation, in this case, with the free bundle.
So, the question we might ask ourselves once we reach the limit of which model we can use is eliminated, but I'm going to change that question for you, and I'd like you to tell me what level you think you have in this area. To answer this properly, I'll leave a link in the description to this free page where you can clearly see where you stand in artificial intelligence. To do this, simply click this button to start a diagnostic.
You'll have to answer some questions that we'll ask you here. In this case, I'll answer them all randomly so we can really see what we'll find at the end of this diagnostic. And when you click on "view diagnostic," we can see how everything is actually being loaded, and it will tell you where you stand.
In this case, according to this response, I would be in explorer mode with a score of 22 out of 100. Some important recommendations and key points to consider include the identified time, the greatest opportunity we have, and the main obstacle we encounter; where we are actually wasting the hours we could automate; important aspects to consider; and even different approaches tailored to your specific situation . This will also recommend different training options, such as our intensive course where you will participate in a video call with me in a group of up to 25 people.
We're starting this course very soon, as we will be closing registration this month. So, take a look at everything, and once we understand this, let's move on to the next section. [music] In this fifth section, we'll quickly see how we can create our own combinations of artificial intelligence models to interconnect them with any tool.
To do this, going back to Omniir Router, if we go back to the Combos functionality directly here, beyond reusing any pre-selected options we have , we can create our own. Imagine you only want to use free models from Antropic or other companies like Deepsic; you can configure this right here. If we click on "Create New Combo" here, we can select everything precisely.
The first thing we need to do is choose a name. In this case, I've called it "Optimized Combo. " They also recommend pre-selecting a free template, such as the " Free Models" template.
However, since we're starting from scratch, I'll leave it without a template and click "Next" so we can choose the specific providers we want to use. To do this, we always have to click on this selector, choose a model we've installed, such as the Nvidia model. From there, we can select from all the models that this provider offers.
In this case, since I'm interested in the free ones, I'm going to select the Deepsic one from here with my Tuto account. With this, I would simply click "add" and this model would be available here. I could repeat this process with other providers.
For example, Orc Router also had a free model, so I'll look for it from here. Here, in fact, we can find the V4 Pro Flash, but I'm going to directly select the one that says "free. " So, let's select the V4 Flash, click "add," and that way I can add all the free models from different providers when I want to use a specific model.
In this case, I could rename the combination to Deepsic Free. For this example, I'll just leave these two models for the sake of the idea. So, we click "next.
" And once we're in this section, we can select a routing strategy, either by priority order, which we can set ourselves, since we can actually reorganize the models, or we can use a weighted method, round robbing, or any other type of routing. And in this case, I'm going to leave the priority setting, and if I click "Next," we can see a summary directly from here where we'll find the name Dieps Free with a priority system, and in which we have two models with two different providers linked. The highest priority would be the Nvidia model, and second would be the Orcast Router model.
Both would use Deepsic Flash. And if I click "Create Combination" now, we'll have it ready to use. Now, instead of using, for example, the Best Free model, we could use Deepsic Free, Cloud Free, OpenAI Free, and, in short, any combination of providers and models that we'd like to use.
In this sixth section, we're going to see another way we could interconnect ONIP Router to other open-source platforms, such as Hermes Agent, which is another tool that's truly amazing, and which I'll discuss in detail in this video. So, now that we know which tool we're talking about and how to install it, let's see how we can link it to free models. Keep in mind that Hermes is very similar to any other open-source tool, such as Deepsic Harness, which we discussed recently in one of our videos.
And we can actually connect it here as well. To do this, if we go to the settings section in the upper right corner and select the providers section, we can click on custom ends. Once we click on it, we simply need to enter the URL where we have our Omni Router tool.
In fact, if we open the terminal that's currently running, we can also see it from here or by clicking on endpoints. It will appear the same way. After pasting this, we just need to give it a name.
In this case, I'm going to call it OVNIRuter, and clicking on test tells us that it has found more than 1700 models, which are the ones we've installed from different providers. So, now that we have all this, notice that we also see different combos here. We could delete the one we have and select any other from here.
In this case, I'm going to select the "best free" one again. And if we click on "save," then we can use it. Notice that in the new session, if I click on this section here, in addition to the previous models I had, I also find all the combinations.
Now, if I click on "best free" again, I can use it here. And if I say anything, like "hello, how are you? " and click on "send," it would have processed this here, telling me "hello Alej, that you're very well," and by the way, if you're wondering, it knows my name because Hermes is one of the best tools in terms of consistent memory.
Again, the video I recommended earlier already discussed this in detail. And if we go to the registration section, So, you can see from here how it appears that it was using the Mimo model in this case, that is, Xiaomi's free model through the Open Code provider, which in fact didn't even require authentication. Here we can see all the details, and notice how it even switched directly to another model, since it first tried Gemini 3.
6 Flash through Antigravity. It encountered an error there, and right after seeing that error, it switched to the next model, which worked correctly. That's why we didn't even notice, but the response to the question we asked worked.
In this seventh and penultimate use case, we'll also see how we can interconnect it with any automation tool via API, since we can also create our own free APIs from Omniir Router to reuse this technology. To see this, I just accessed an empty flow within my N8N platform, which I have on a VPS. And from here, notice that if I click the plus sign and link it to an AI agent in the AI models section, I'll actually be able to link all these free providers.
But what's most interesting is that I could connect it to our own OmniRouter, which actually encompasses many of these. To do this, all we have to do is click on the OpenAI API, since the OmniRouter API is compatible with the OpenAI API. And in the credentials section, we click on "create credential.
" So now, since it asks for an API key, I'm going to open another OmniRouter that I have directly on a VPS. I'll explain that in more detail in the last use case. And if I click on "API key manager," notice that from here we can create our own key.
I'll click on it. Notice that here we can also enable or disable different functions that this API key we're going to create can perform. And once we have these configurations that we want to activate or deactivate, after giving them a name, like in my case where I'm going to use N8Ni, creating the API key, we'll find this one here that we can copy and paste into our N8N.
Now, regarding the URL base, instead of using Openi's, we would use the one we have on our end. In this case, since I have it on a public cloud, which would be this URL here, I'm going to copy it, paste it from here, and if I click save, it tells me that everything has been configured successfully . Now, if I close this and go to the models section, I'll find a lot of models beyond Openi's, such as the combo boxes we have like the Best Free one, as well as other models from all the providers we've interconnected.
[music] Let's finish with the last section by taking a general look at all the functionalities we have in OVNIRuter, without going into too much technical detail. To begin, keep in mind that by going to the Omniproxy section, you'll find all the settings related to AI models and providers. We've been talking about the NPINs we've used, creating Happy keys, and the providers we can link, but here we also have so many combinations, a combo box for a more professional and automated approach, and even the ability to see the share of each provider.
In addition to this, we could also go into detail about the context understanding section . From here, we can access each of these options, but we'll only cover them briefly, as I don't want this to become a very technical video. However, if you're interested, let me know in the comments if a future video on this topic would be worthwhile.
Besides all these options, which, as you can see, are numerous, we also have the analytics section. And from here, beyond viewing the logs, which is what we've been testing, we can also easily see the activities. We have all the logs and the usage timeline we've been working on.
As you can see, this would be active every second. We could also audit the log, the MCPs, and perform a 2A audit. And we'd be able to see the health of our system at a glance.
Along with this, we'll also have different developer tools, such as a playground, which is very useful for testing all the linked models, like chat models. If I select a provider like Grock and choose one of their free models, such as OpenAI, and we type something like "Tell me an interesting fact about your model" and click "Send," it could automatically provide the response, as you just saw. From here, we could also manage the system distribution and all these parameters, allowing us to test each provider and model directly from this interface before integrating them into our own systems.
Keep in mind that we also have some additional options here, such as the ability to compare prompts, allowing us to run the same prompt on different models and compare different results for a much more visual overview. Beyond all these tools, we also have some agent-based functionalities and, ultimately, a whole host of options to customize and configure the platform to our liking. Another important point about this tool is that it's open source, as I mentioned at the beginning of this video, so in addition to having it on your computer, you could have it on your own server.
I, in fact, have it on a VPS within my Hostinger account. Here, as you can see, I have the Omnir Routro prompt, which I can open directly from here. The great thing about this is that it would be available 24/7 without having to rely on my own computer, since I've seen that it's also available directly from Hostinger's Docker catalog with just a click.
I'll also be leaving this link below with a coupon I have, but keep in mind that this video isn't sponsored by them or anything, but I personally find it really interesting and worth considering. The truth is, with this we can access almost everything, although perhaps the real advantage isn't having access to everything, but being able to find a route when it seemed like there were none left.
