Hermes + North Mini Code: New FREE API!

**Julian Goldie** (0:00)
There's a brand new free API and free local model that you can use with Hermes Agent, and it's called North Mini. I'm gonna show you exactly how it works. We've run it both locally and also with the API, pretty powerful stuff, super fast, and it's designed for agentic coding. We've already plugged it into Hermes Agent, like you can see, and we said, like, you're working, and it actually replies, like you can see right here. So you can work with Hermes Agent, it's pretty quick to reply, and you can do all sorts of stuff. So if we say, for example, okay, schedule a Japanese practice session at 5 a.m. today or 5 a.m. tomorrow, plus daily as a scheduled task inside Hermes, we can plug it in, and then Hermes will use North Mini to just go off and build this. Now, you might say at this point, okay, how does this work? What are the benchmarks like? How powerful is it? How do you access? I'm gonna cover all of those questions today. So number one, how do you get access to it? So you can actually get access on Open Router. So if you type in North Mini, you'll see North Mini code is ready to go. It's 256K context window, and it's free to use, and you can plug it into Hermes Agent directly. Pretty awesome. Now, how do you actually set this up with Hermes? All you do is you go to your terminal like so, and then you go to Hermes model, select Open Router from the list, and then you would select North Mini code as the model for your Hermes Agent. That's how easy it is. So just type in Hermes model, set up Open Router.
So what is the API? It's this free one. How do you get access to it? Inside Terminal. Now, you can also run it with Ollama. So if you prefer to use Ollama and run it locally, this is not a cloud model from what I've seen. So you can install Ollama in one single click using this terminal command. And then from here, you just go to this section, and you can see again, it's a local model. It's not cloud-based, which means it's free for anyone to use and run locally. And then to pull the model, you would click this, paste that into your terminal with Ollama running. To run it inside Hermes Agent, you can use this terminal command. You can also run it with OpenClaw, Claude Code, Codex App, Codex and OpenCode as well. Pretty awesome. And so if you are out of tokens, or if you've run out of usage on your existing limits, you can just use a free model like this and plug it into Hermes, right? Nothing can stop you. That's pretty cool. And so this is how you can use the system, how it works, etc. And this is something I call the Free API Command Engine. So you can wire it into Hermes and you get an agent that can write files, run tools, build real things for you for free. And just for fun, we built this out with Hermes, but you can use it agentically and it can call tools. It's actually a very small model.
So this is a really small model if you're running it locally, which means it's really fast and it outperforms Gemma 4 on some of the benchmarks. It's 30 billion parameters with 3 billion active. It's one command to run it, 250, 6K context window, and free whether you use it on the API or if you use it directly with Ollama as well. So you got two different options right there.
And you can see that it's working, it's scheduled the task, pretty simple and easy. And you can see that it's currently scheduled that Japanese practice that we just talked about right there. Pretty nice. So here's the announcement of North Mini and its setup. And you can see how it performs on benchmarks. So again, you would compare it against models like Quen 3.6, Gemma 4 These are the comparable models. If you want to understand how it works, et cetera.
And you can see the quote from Cohere here who built it. So North Mini code is Cohere's first agentic coding model, a 30 billion parameter mixture of experts model with 3 billion active, optimized for code generation, agentic software engineering and internal tasks. That is perfect for Hermes Agent. It's also available on Hugging Face and OpenRooter for free as well. So you can see the details on Hugging Face right here. In terms of the architecture, how it works, everything else. So how can you wire it into Hermes? You just grab a free OpenRooter key, create a Hermes profile for it.