I Cancelled Every AI Subscription and Set Up 14 Free AI Models on My PC | FreeLLMAPI Tutorial

Yo, what is good everybody? Welcome back

to the channel. So, real quick, raise your hand if you've ever wanted to use AI tools without worrying about monthly costs. Chat GPT Plus, Claude Pro, Gemini

costs. Chat GPT Plus, Claude Pro, Gemini Advanced, it adds up fast, right? What

if I told you you can access all of this using the official free tiers these companies already provide? All from your own computer, all at the same time. I'm

talking Google Gemini, Grok, Cerebrus, Nvidia, Mistral, GitHub models, Hugging Face, and like 10 more on top of that.

You can access all of this using the official free tiers these companies already provide. All right here on your

already provide. All right here on your Windows machine. Today, I'm going to

Windows machine. Today, I'm going to walk you through setting up free LLM API, a self-hosted open-source tool that stacks all those free tiers together and gives you one single endpoint to rule

them all. It's honestly one of the

them all. It's honestly one of the coolest projects I've come across in a minute. And I promise, even if you have

minute. And I promise, even if you have never touched a terminal in your life.

All right, before we install anything, let me explain what free LLM API actually is. Because once you get it,

actually is. Because once you get it, you're going to be like, "Why didn't somebody do this sooner?"

So, here's the deal. Almost every major AI lab, Google, Meta, Mistral, Nvidia, they all offer a free tier on their API.

That's usually a few million tokens per month, a few thousand requests per day, for free.

The problem? Each one has its own SDK, its own rate limits, its own way of doing things.

Managing 14 different accounts and 14 different APIs can get complicated really fast. Free LLM API solves that.

really fast. Free LLM API solves that.

It sits on your machine as a local server, and it wraps all those providers behind one single OpenAI compatible endpoint. You just point your app at

endpoint. You just point your app at localhost. The router automatically

localhost. The router automatically picks the best available model, manages your requests efficiently, and moves to the next provider if one is unavailable.

We are talking hundreds of millions of tokens every single month. On top of that, you get a built-in admin dashboard, so you can manage your API keys, watch your analytics, test prompts

in a playground, and control the fallback chain, all from a clean web UI.

No command line required after setup.

And because it's self-hosted, everything runs locally on your own Windows PC. No

third-party relay involved. Okay, let's

get your machine ready. We only need two things installed before we start.

Node.js version 20 or higher and Git.

That's it.

Step one, install Node.js.

Open your browser and go to nodejs.org.

You'll see a big button that says get Node.js. Click that. On the download

Node.js. Click that. On the download page, click on Windows installer and it'll download an installer.

Just double-click it and hit next through the whole thing. Default

settings are totally fine.

Once that's done, let's verify it worked. Hit the Windows key, type cmd,

worked. Hit the Windows key, type cmd, and press enter to open command prompt.

Now type this exactly. node {dash}

{dash} version Hit enter. If you see something like v24.something,

you're good. If you get an error, close the command prompt window, reopen it, and try again.

Sometimes it just needs a fresh start.

Step two, install Git.

Now go to git-scm.com and click install for Windows and download the installer.

Same deal. Run the installer, hit next through everything. Default settings all

through everything. Default settings all the way.

When it's done, back in your command prompt, type git {dash} {dash} version.

Hit enter.

You should see something like Git version 2.something. Perfect. We're

version 2.something. Perfect. We're

ready to rock.

All right. Now, let's grab the actual project. In your command prompt,

project. In your command prompt, navigate to wherever you want to put this. I'll just navigate to my C drive

this. I'll just navigate to my C drive in Windows Explorer and open it in terminal.

Now, clone the repository. Type the

command displaying the screen or copy it directly from the video description. Hit

enter. You'll see it downloading files.

That's normal. Just let it finish.

Now, move into the folder. CD free LLM API.

Hit enter.

And now, install all the dependencies.

Type npm install. Hit enter. This one

might take a minute or two. It's

downloading all the packages the project needs. You'll see a bunch of stuff

needs. You'll see a bunch of stuff scrolling by. That's totally normal.

scrolling by. That's totally normal.

Just let it run until you get your cursor back. If you see any errors,

cursor back. If you see any errors, these might be due to some protections are enabled or some warnings. You can

simply try the command npm install --force and it should install the dependency packages.

[clears throat] Okay.

This is the one slightly technical step, but I promise it's super easy.

We need to create a .env file that stores your encryption key.

This is what keeps your API keys safe on disk.

Still in your command prompt, type the command on the screen or copy it from the video description.

Hit enter.

Now, we need to add an encryption key to that file. Copy the whole command and

that file. Copy the whole command and hit enter.

If you see the error like this, you can try replacing the percentage symbol with the dollar and bracket. I have added both versions in the description of this video.

So, it worked for me without the percentage symbol.

What that does is generate a secure set up key that protects your stored credentials. Your credentials are stored

credentials. Your credentials are stored securely on your own machine.

Pretty solid for an open source tool. To

double-check it worked, you can open the .env file in Notepad.

.env file in Notepad.

Type notepad.env.

Hit enter.

You should see the encryption_key with long random strings.

Close Notepad and we're good.

All right, moment of truth.

Let's fire this thing up.

In PowerShell, type npm run dev.

Hit enter.

You'll see the terminal spring to life.

The server boots up, Vite spins up the dashboard, and within a few seconds you'll see something like localhost:5173 in the output.

You can press control and click to follow the link.

And there it is. Your own personal AI dashboard running right here on your machine. No monthly fee. Nothing.

machine. No monthly fee. Nothing.

Okay, this is the fun part. We're going

to add our free API keys.

Click on the keys tab in the dashboard.

Now, all of these providers are free to sign up for.

Let me walk you through the first few so you get the pattern.

Let's start with Google.

Go to ai.studio.google.com.

Sign in with your Google account.

Click get API key.

Click create API key. I'll name it as free-llm-api.

And in here, you can choose any existing project or also create a new project.

Copy it.

Back in the dashboard, navigate to keys tab.

Choose Google AI Studio from the drop-down. Paste the API key and click

drop-down. Paste the API key and click add key.

Done.

You just unlocked Gemini 2.5 for free.

Then, go to console.groq.com.

Sign up for free.

Go to API keys.

Click create API key.

Copy it.

Back in the dashboard, select provider Grok.

Paste the key and click add key. That

gets you Llama 4, Quinn, and Kimmy running at some of the fastest inference speeds you'll ever see. Grok is

genuinely one of the fastest inference providers out there. And here you can see the Google key is checked and showing as healthy.

GitHub also offers an official free developer tier. If you already have a

developer tier. If you already have a GitHub account, you can sign up and get access to models including GPT-4o.

Go to github.com.

Click your profile picture, settings, developer settings, personal access tokens.

Generate new token.

Give it a name, set it to no expiration, and hit generate.

Copy that token.

Add it in the dashboard under GitHub models.

The same process applies for every other provider. Cerebras at cloud.cerebras.ai,

provider. Cerebras at cloud.cerebras.ai,

Nvidia at build.nvidia.com,

Mistral at console.mistral.ai,

OpenRouter at openrouter.ai, Hugging have free tiers. They all take about 2 minutes to sign up, and they all work

exactly the same way in the dashboard.

Sign up, get key, paste key, save.

Repeat.

The more keys you add, the more providers you have available, and the more smoothly everything runs. Now,

click on fallback chain in the dashboard.

This is where you tell which provider to try first, second, third, and so on.

I recommend putting Gemini 2.5 at the top. It's the most capable free model in

top. It's the most capable free model in the list right now.

Then Grok for speed. Then GitHub models for GPT-4o.

Then the rest in whatever order you like. The router automatically picks the

like. The router automatically picks the best available model and distributes your requests across providers. When one

provider is busy or unavailable, it automatically moves your request to the next one in line. You don't have to do anything. It just works.

anything. It just works.

Click back to the keys tab and look at the top of the page.

You'll see your unified key. Copy that.

This is the one key you use for everything. Want to use it with any

everything. Want to use it with any OpenAI compatible app? Just change the base URL and paste this key. That's it.

It works with LangChain, Llama Index, Open Web UI, anything that supports OpenAI's API format. And here's a cool thing. Click the playground tab. You can

thing. Click the playground tab. You can

send test messages right from the browser. Pick model auto. Let the router

browser. Pick model auto. Let the router decide. Type your message and hit send.

decide. Type your message and hit send.

You'll see the response come back and right on the message it shows you exactly which provider served it and how long it took. Super transparent. Let me

change the model to Gemini 2.5 Flash and give it a prompt.

What is LangChain? Here is the response.

So, as you can see the provider, the model, and the time it took to respond is showing here.

So, use Grok if you need blazing speed and set the other on fallback.

Let me just real quick run through why I think this project is genuinely awesome.

First, it's self-hosted.

Your API keys never leave your machine.

You're not trusting a third-party relay.

Full privacy, full control. Second,

800 million tokens a month, combined, for free.

That is more than most individual developers will ever burn through in a month. You can build side projects, run

month. You can build side projects, run experiments, learn prompt engineering.

Third, OpenAI compatible.

You can literally swap into any app that already uses the OpenAI SDK.

Just change the base URL, one line of code. Fourth, automatic failover.

code. Fourth, automatic failover.

If one provider goes down or hits rate limits, the router silently falls to the next one. Your app never knows the

next one. Your app never knows the difference. And fifth, that dashboard is

difference. And fifth, that dashboard is clean.

Key management, analytics, playground, fallback chain editor, it's all right there in a nice dark mode UI. No extra

config files, no YAML, nothing complicated. And that's it, fam.

complicated. And that's it, fam.

You just set up your own personal AI server on Windows. 14 providers,

hundreds of millions of free tokens, one clean dashboard. The GitHub link is in

clean dashboard. The GitHub link is in the description. Go give it a star. The

the description. Go give it a star. The

dev deserves it. And drop a comment below and let me know which providers you're adding first. I'm curious what y'all are building. If this video helped you out, smash that like button. It

genuinely helps the channel grow so I can keep putting out free content like this. And if you're not subscribed yet,

this. And if you're not subscribed yet, hit subscribe.

I put out AI tutorials, free tools, and side project builds every week. I'll

catch you in the next one. Peace.