Qwythos 9B is INSANE (FREE + Local + Open Source)!

We're going to be looking at Quifos 9b, which is a new local model. As you can see right here, it's come directly from, it's available on Olama to host, and there's many different models. You'll see the context window here is 1 million token context window. I'll come on to that in a second, but essentially this is designed to be like a Claude style creative reasoning model with a Quen 3.5 9b base.

And so what you can do here is you can easily install it from Olama. If you don't have Olama set up already, this is a way to run free models. So you can use this terminal command. Once you've done that, make sure you have Olama running.

Then you can run this terminal command and you can download the model and start running it directly. We've already plugged it into our agent OS system so that we can test it out. And you see, for example, we've got it inside the build section here. I'll show you what we've built within a second, but we can see all of our creations inside the local engine of our agentic OS.

So I'm going to run you through what it means, how it works, what it is, why it's different. And this is basically a free Claude style AI that runs on your own Mac. Now, is it like OPA's 4.8 level? Probably not, probably not, but is it decent and can actually create stuff locally?

Absolutely. So QueenFos 9b is a Claude Miphos style creative reasoning model that runs 100% on your machine, which means no cloud, no token costs, nothing leaving your computer. And we've already wired it into the agent OS local engine, tested it out, and I'll show you exactly how it works. So you can see the details here.

So it's pretty small as a model, only 5.6 gigabytes when we ran this version and it was free to get from Olama. You'll also see some stuff that we created with this just for fun, you know, to test out, see how it works. So we create like a little to-do list here. As you can see, we can add new tasks.

We can clear everything that we've created. So you can create like little mini apps. This was like a digital clock that we created that runs locally. And this is like a little snake game, classic, classic.

So we put it to work. We tested it out. It actually works, which is pretty cool. And it's running privately.

Here's a few more things. So actually one thing that surprised me genuinely was that it could create like a fairly nice design. I've seen, I've seen like frontier models create worse stuff than this for the landing page, which was pretty nice. It created some stuff that was a little bit glitchy, but it even created like this calculator that seems to work as well.

So we tested it with like six apps, all three, running it with Olama. You can see how it works inside the terminal right here as well. And there's two models that I've seen recently that have been interesting for local, which is Qwitfos and also Ornith 9b as well, which is something we tested out from what I can see so far. I would genuinely say that probably Ornith 1.0 is outperforming it.

I've got another tutorial on that. But so far it's pretty good. Now, if you're wondering, okay, what's the build speed like? So you can see the Qwitfos was a lot slower.

Sorry, it had more tokens per second. On short and build designs, it's a lot smaller as a task. It can build working apps. And the context window is insane for this, which is 1 million tokens.

And also, you know, Ornith and Qwitfos are both local free models. Now, both of them are great. You know, but for the same work, Qwitfos actually came back twice as fast, which is pretty cool. Now you might be wondering, okay, why is it called that?

So it's a stack of four deliberate moves laid on top of an open Qwen 3.5 9b base. That's why a model small enough to run on a laptop. I mean, five gigabytes is really, really small for a local model. It can write a reason like something far bigger.

So the way this works is like, it's based on Qwen 3.5 9b, which is a strong open 9 billion parameter base model. Then you've got something that's trained on like Claude style reasoning and creative. From there, it has a million token context window. So the context window has been stretched and it's also quantized with Ollama and Llama CPP.

So it can think like Claude locally. It's designed for creative stuff. Big context window, reasons and calls tools, which is good for agentic. You can use it like so.

You can all see the benchmarks here as well. Also, you might say, okay, million token context window. That's amazing. That is really the ceiling.

It's not going to run out. It's not going to run like that out of the box. So for example, it can still run out of tokens on requests because Ollama would load it with a much smaller window. The token window was tiny when we tested it out, actually.

So depending on what memory you have, that's going to determine how many tokens this actually runs with. So the longer the window, the bigger the memory cost. So in the real world, you would pick a window that fits your RAM. And then you can see the different models here.

So we use this model. There are some higher quality, more heavy models here, as you can see. And you've also got a near lossless state for Q8 as well. So there's lots of different models in terms of the size and everything else from Ollama, so you can go with smaller, which is lighter and faster.

But slightly less sharp, or you can go with bigger, which is closer to the original. So that's why we ran it in the middle. Honestly, I've run it on a Mac studio. You can see my settings over here.

And it does struggle with local models sometimes. So I still don't think like local models are anywhere near frontier level, but they're pretty fun to try out and they are getting better and better. So what's great about this, it's free and it's private. It's pretty fast for its size.

It's pretty light, Claude style brain, agent ready. So it has native tool and function calling. And it can actually build stuff. What's not great about it, obviously it's an IB, so it's not a frontier model.

The million token context window doesn't actually happen in reality based on my personal setup. It can be a bit slow loading up and there's no built-in tools or memory or anything like that. It's just a model. Occasionally it gets a bit messy as well.

When we tested it, sometimes it would work really well and then sometimes it'd be a bit messy. So for example, like I was talking about the context window here, when we were testing out inside the chat, sometimes it wouldn't fully load or it just cut off as you can see right here, but then sometimes it would work perfectly as you can see right here, right? So sometimes it does get cut off despite the large context window. So overall it's pretty cool.

Cool. You know, it's worth checking out. Quifos 9B, 9 billion parameter built on Quen 3.5, Claude style thinking, thinking model, million token context window. You might say, well, it can't be any good, but actually you've seen it run.

You've seen some of the stuff we've created. It's not bad. How it works is you ask it, Alarma runs it, Quifos thinks, and this all stays offline on your Mac. So if you don't even have internet, you can still use this AI, which is pretty cool as well.

You also might think like, I need a massive setup for this, but it's only 5.6 gigabytes. To put that in perspective, like if you're installing Gemma 12B, that's about 20 gigabytes. So it's about a quarter of the size of Gemma for 12B, for example. Bear in mind as well, you can plug this into whatever you want.

So whatever agent harness you want. So you could use this, for example, locally via Alarma with Hermes agent as well. I haven't tested that. I don't know if you'd get good responses, but it could be worth testing out and seeing for yourself what it's like.

And if you want to get our local engine wired into a system, not just a chat with the terminal, then, you know, a local model is powerful, but a local model running as the engine of your whole agent OS is pretty much unstoppable. So inside the Airprofit boardroom, you get our agent operating system that wires Quifos in as the local engine. So your agents can run on it by default. That means it's free.

It's private. It can run with whatever you want. And the full agent operating system, including the local engine setup, the every prompt, the dashboard for live coaching calls a week, daily tutorials, token optimization tutorials, et cetera, and a community of 3,800 builders. That's all inside the AI profit boardroom.

Link in the comments description or go to the airprofitboardroom.com. If you're wondering what you get with the agent OS, you can see some examples here. So we have the local engine set up, which means that we can use local models. We can switch this in and switch this out.

So every time I test a new model, I switch the model and then we can then see the workspace here. So everything that we've built, we can preview everything that we've created as well, which is pretty cool. We have free cloud code as well, which you can run with local models. And then for example, if you like agents like Hermes, we've got all these different styles of using Hermes.

So we have Hermes Jarvis, for example, Hermes Oracle built in for the latest news, a studio, we've got a workspace for seeing everything we've created, a lead generation tool as well for outreach. And then we also have mixture of agents, which is a brand new model from Hermes, a new system you can use to get like better outputs than Opus 4.8. It's pretty wild. So if you want to get that link in the comments description or go to the AI Profit Boardroom, this is my AI community for helping you save time and grow with AI automation.

Lots of people inside there using local models. Inside the classroom, you get access to all of our best trainings and you can get the agent OS from this section along with new daily video tutorials and guides based on what you want to use. Inside the calendar, you can jump on weekly coaching calls, get help and support in real time. And inside the map, you can meet people in your local area who are building with AI agents like you.

So feel free to get that link in the comments description or just go to the AIprofitboardroom.com. Thanks for watching.