New FREE Hermes Agent Model!

Julian Goldie (0:00)

MIMO V26 from Xiaomi just dropped. And this is a powerful new model, it's actually frontier level, it's open source. And not only that, but also you can use it for free and you can use it inside the Hermes agents. So Xiaomi just dropped MIMO V26. It is the strongest open source model ever released. It has two different versions, a Trillion Parameter Pro that sits at one point by GPT 5.6 Sol and a Flash version that is a lot cheaper and faster as well. So I'm going to show you how to plug both into Hermes and the HNOS in about 10 minutes. Your agents get Frontier Brain and you can use the Flash version for free as well. So let's get straight into this. If we actually look at the numbers, you might be saying, okay, Xiaomi, MIMO, is that actually any good? So if we look at the Artificial Intelligence Analysis Index, as you can see right here, it's called 46 on the Artificial Intelligence Index, the highest any open weight model has ever reached.

Now, just because it's open source doesn't mean you can run it locally, like it's, you know, a super heavyweight. But if we look, for example, here, it's just behind MIMO v2.6, actually built this 3D game, as you can see right here. And we've actually got inside the HNOS, a separate profile built for MIMO Flash, as you can see. Now, you can also get the Open Router APIs. We'll unplug that into your HNOS system, but this is how you can use it directly. And then you've got two versions. See, you've got MIMO v2.6 Pro, which is a trillion parameter model, and you have MIMO v2.6 Flash, which is a 309 billion parameter model as well. You see another example of what it built over here. So it can build like 3D games and that sort of thing as well. It's also pretty good at computer use. So for example, if we test it out now, and we'll test it with MIMO Flash. So we're going to get this guide, for example, here. And then if we go back inside the chat, we'll do forward slash learn, and then plug that in. And we'll just test how fast it is, and if it can actually run agentic tools, and how powerful it is with Hermes as well. Now, interestingly, on agent work, and this is what it's really designed for, which is why it's so good to plug into MIMO Flash, it ties or beats Claude Opus 5 in most rooms. So if we look at the numbers over here, you can see that it's right up there with DeepSWE, with Terminal Bench 2.1, which is computer use as well.

We've got Automation Bench, and we've got Agents Last Exam as well.

Now, if we go inside a terminal, you might be wondering, how do you set this up? So I would recommend, if you want to use this for free, here's how you can do it. So you can go to Hermes Setup, like so, and then over here, you can switch over to OpenCode. Now with OpenCode, you can use OpenCode, and you can select the model as MIMO Flash. So if we go over to OpenCode here, you can see that we have MIMO V26 Flash 3

That is a model that will have plugged into OpenCode. So you can use that inside OpenCode for free, but then you can also plug OpenCode inside Hermes Agent as well, which is pretty nice. Now, if we look at the prices here per million output tokens, you can see MIMO V26 Flash is a lot cheaper, but again, you can use it for free inside OpenCode. MIMO V26 Pro is $0.87, which is pretty cheap for a frontier level model. And then you got Pro Ultra Speed, which is 20 times faster. And you can see it is quite an expensive model to use. I don't recommend using that unless you need like really, really fast answers. This is also a mixture of experts model. So you can plug in a topic and then it will route that to 8 of 256 billion parameters as it's going through. Sorry, 256 like expert models inside there. And then you can see here that it's used the skill.

Now we'll say it's still not that fast when you're using it. This isn't like a super fast model. Don't be, I would say keep your expectations low, even if it says flash on the end of the model name. So if we type in, for example, forward slash learn, and then the guide over here, it saves the skill, but it took about, I would say 100 seconds to come back to us on that, to read the guide to create the skill inside Hermes agent. So it's good, but it's still not that fast. But at the same time, this is a frontier level brain that you can plug in for free into Hermes agent. So this is how we created the separate profile for MIMO inside the agent OS. So you can just type in Hermes profile, create MIMO-pro, and then create the new profile like so. And we've got MIMO v2.6 pro inside there as well. And then from there, you just want to test it and make sure it actually works. But you can see some examples of us using it inside the agent OS over here. And it works pretty nicely. So pretty cool. By the way, Xiaomi from China, so worth knowing as well. It's a Chinese model. And also they're releasing like loads of new models recently. So for example, v2 pro only came out in March, but they're on v2.6 already. So April, they released v2.5, and then September, they released v2.6.