Hello everyone, Pride here again, this time with something I wanted to do a lot on this site.

You, my dear intermediate user of our glorious J.ai, must have already come across a message in the bot saying "Proxy On" or "Off". But since you are just a normal person who wants to with the bot, you never stopped to think about it.

But...to be honest, what is a proxy?

Proxy:

Proxy is not an exact thing, it can be used for many things on the web as a support for certain things, such as: Firewalls for websites, assistants and finally LLM models.

Anything here on Janitor except for Open AI and Janitor LLM is a proxy.

With a proxy, you can use several paid and free models to be able to chat with your bot and have a different experience than what you are used to with the standard model that we have. And no, I'm not saying that the JLLM model is bad, that's up to anyone's personal taste.

But enough nonsense and let's get to what you want most.

Openrouter:

Openrouter is very simple to explain. It's a website that you can use to get these models, every week (I think) new models always appear and they are divided by the month in which they were released.

ᅠᅠ

(Openrouter for case you want to see)

ᅠᅠ

Now let's go step by step on how to get a Proxy model for yourself. And remember that not all of them are free, and at the end, I will leave a small list of the best (in my opinion) in each category.

After entering the website, just create an account by clicking/pressing on the top right side of your screen to create your account for openrouter.

After that, you will follow a list of simple steps on how to create a key and also choose your model for your proxy.

• Step 1: The first step is to go to the website settings, which is very simple. Go to the top right corner of the screen and click/press the button with three bars: Image example

• Step 2: After that, you click/press on settings: Image example

• Step 3: Once you get to the settings, scroll down to the bottom and you'll find this

• Step 4: Choose your preferred model, normally, the free ones always have (free) at the end.

• Bonus step: In case you want to have the model of a certain model and you don't know how to get, click on these two links below and see how to get a model whatever you want.

https://files.catbox.moe/i2uz3o.jpg

ᅠᅠ

https://files.catbox.moe/wgt2s9.jpg

Now let's go to a really quick one, which is the keys.

Still in the top right corner of the screen, you go to Keys and there you will be able to create your key. It is a very simple process and without much fuss. You just enter the name of your key and the limit it will have (the limit part is completely optional).

Now, finally with your key in hand, before I have to deal with something: https://openrouter.ai/api/v1/chat/completions

This is not a link to use, you will copy this 'cause it's important.

ᅠᅠ

Now follow my steps. First you will enter a chat, preferably one with the active proxy. After that, you press the purple button on the top right side of the screen and go to proxy.

First: Put it in a custom model so you can place your proxy model, and then paste the model you have.

After that, you will use that link you copied here and click on check the API model.

Image example

ᅠᅠ

And after that it's very simple. Just go a little bit down and paste your key and press save settings and be happy with your bots now with a different model.

Models.

The name is self-explanatory, I will talk about models that I used and liked and I want to share them with you in case you are interested.

• 1: Deepseek: Many people's favorite, it's the one that always gets the most praise and is always talked about on this site's Reddit, but is it really all that? Deepseek has several models; some are free and some are not, but the ones most people use are the Deepseek R1 and V3. They are both quite different, so I will try to summarize each of them.

• R1: This is the most used of all, it's what made a lot of people want to use more proxies and beg bot creators to activate the proxies (I think someone does this...) But Deepseek is really everything they say: It's a model with perfect memory and very well programmed context. The model makes the bot's personality very faithful to the character you're using, making a huge difference from what is normally seen in the site's standard LLM, which is sometimes a bit cheesy and too romantic. But, not everything is perfect because using Deepseek is very annoying! First, the characters usually become very cynical and sarcastic, like the creature from Brain Damage (nobody knows that movie, just me...) and that's really annoying in certain bots, and even though they're very good at describing, sometimes the model says something completely unnecessary, like: In the distance, in the big city traffic, Pablo trips over his own foot. Why the hell do I want to know that!? And also this model is a ass for most MHA bots, seriously, don't use this model if you want to have a normal chat with some character that has powers, because every time it will describe something completely unnecessary and useless about how such character is doing such thing because their power does that: I DON'T CARE! I JUST WANT TO HAVE WITH MIRKO!

ᅠᅠ

• V3: This one is very simple, it is essentially the Janitor but with improved memory and also with slightly worse attributes of the R1, but even so it is a good model and I normally use it more than the R1.

  ᅠᅠ

Modela for both: deepseek/deepseek-r1:free + deepseek/deepseek-chat:free

• Google Gemma 3: I think this is the one I have the least to say about. It's only good for SFW scenarios, for something more NSFW this model is shit.

ᅠᅠ

Model that I use: google/gemma-3-27b-it:free

And that's it, I hope you enjoyed this lazy walkthrough. It was really fun to make and I hope it helps a lot of people. Have a good day, afternoon and evening.

I little (late) update: Openrouter has now a limit of 50 messages. If your country isn’t shit like mine, you can pay 10$ for 1000 messages.