China’s Minimax M3 + Hermes AI Agent is INSANE!

Added: 2026-06-02

[00:00:00]MiniMax M3 just dropped out of China.

[00:00:03]When you plug it into Hermes agent, you get a free AI worker that can run on your computer for 12 hours straight.

[00:00:10]We're talking about an AI that can open up your app, schedule tasks, search your web, send reports all on its own using this powerful agentic model that literally just dropped. And in this video, I'm going to show you how to set up, test it live with real tasks, and there's one thing you absolutely have to do to unlock the full power of this.

[00:00:28]Plus, I'll show you something really cool at the end, which is a cloud-hosted version of this you can access from anywhere. Let's get into it. So, this is MiniMax has just come out from China.

[00:00:39]MiniMax M3, brand new coding and agentic model. And we can check out the benchmarks here as an example. So, you can see here that it's performing pretty well on benchmarks. And what you can actually do is plug it into Hermes using these commands. And you can actually test out for free using these commands, too. So, the way that you can get started with this if you want to plug MiniMax M3 directly into Hermes agent is we can use the commands right here. So, the first thing that we need to do is make sure that we have Hermes agent installed. Make sure that you have Ollama downloaded as well. And once you've done that, make sure you have Ollama running in the background like you can see. Now, we can run this simple command. So, Ollama launch Hermes and then pick the model, which is MiniMax M3, of course. You can run that inside a terminal and boom shakalaka. We now have Hermes agent connected to MiniMax M3 cloud. Perfect. Now, if you're on the free plan on Ollama, obviously there's token limits. Most people won't hit those. But also, the cool thing about this is in one single click, you can also use MiniMax M3 inside Claude code, inside Codex app, inside OpenClaude Codex, OpenCode, etc. So, what you can also see here is it's a 500,000 token context window, and the input is text and image as well. Now, if we go into my Agent OS, you can see that it's automatically synced to MiniMax M3 Cloud, which is perfect. So, if we go and test this out now, we can try it over here. And just for a quick agentic task, we'll say go to chat.openai.com, and let's see how that goes. So, another test I've run here is open up the notes app locally and say hello. And you can see it says note created, and the notes app is now in the background with the note typed in. Here's what it did. And then, if we go over to our notes, we can see hello from Hermes, right? So, it can talk to us through the computer. It can navigate to local files. It can open up apps, and also it can do that in the background whilst we're talking to it directly, which is perfect. Let's say schedule afternoon tea at 3:00 p.m.

[00:02:47]daily, just see how it responds to more agentic tasks to tasks running on a schedule. The one thing that I'll say here is if you are using it through Ollama, it can be a little bit slow because it's running through the cloud models, right? If you want a faster version, what I actually recommend is you either use OpenRouter, or even better, you can actually use M3 directly with the coding plan, and then use that to use Hermes, right? And the good thing about that is once you signed up to the coding plan, then you don't need to use APIs because you can use OAuth to log in to MiniMax directly. Now, also something that's interesting here, and 99% don't even know about this, but you can actually get a cloud-hosted version of Hermes called Max Hermes. And since this release of MiniMax M3, it's now powered by M3 as well. So, if you're using Max Hermes in the cloud, you can deploy it in one single click, and then you can use Hermes and also Max Claw, which is a version of Open Claw in the cloud, too.

[00:03:46]All right, so if you wanted to deploy this, you can just click on start now, log in, and then you just need to make sure you have the plan like so. So, this is what it looks like once it's deployed. Honestly, for me personally, you know, you could use a terminal. Do you have the full conversation history?

[00:03:59]No. You could use Max Hermes in the cloud, but it still doesn't look that nice. I personally prefer using my Agent OS system. It's a lot nicer. It's a lot easier to navigate, and the UI is just beautiful, right? The other cool thing about this is everything that you've created, you can preview live as well, which is pretty cool. So, at this point you might be saying, "Okay, what is MiniMax M3?" Well, the way that I look at it is the MiniMax M3 is the brain, right? So, it's a brand new AI model, and you want to think of it like the thinking engine. So, it's a brain that you can plug into your AI tools, and it can run for 12 hours autonomously on different tasks. It can work on long horizon tasks. And also, it can actually improve itself on its own. It's a self-learning improving model. And it's designed for agentic capabilities. Now, you might say, "Okay, what is Hermes?"

[00:04:46]So, Hermes is the agent that you can plug it into. So, if MiniMax is the brain, then Hermes agent is the body.

[00:04:53]It's the part that can actually do things. It was created by Nous Research, and it's one of the few agents that has an autonomous self-learning loop, which means the more you use it, the better it gets. And so, why is this so good working together? Well, the brain MiniMax M3 is free if you use it on a Llama. It was a long CC and token limits. Or you've got And And the body itself, Hermes agent, is free and open source, too. And so, when you stick them together, you basically have a very smart helper that can write your content. It can search the web. It can send emails and reports. It can control apps locally. It can run tasks on a schedule while you sleep. It can learn your business over time, right? Uh the other cool thing about all of this is you don't need to be up to code to use this. You can just use it inside the chat here. Now, you might say, "Okay, how does it perform on benchmarks?" So, we can see M3 easy as high-performing as Opus 4.7. Not on SWE bench, but it is comparable. And obviously, it's a lot cheaper than using something like Opus 4.7 as an API. The other thing here as well is that it's pretty good for example, genetic tasks who can use MCPs and stuff like that. And also it's an open weights model, too. That's the thing. So for frontier level, they've actually said they'll soon be fully open sourced on Hugging Face, which is pretty cool, too. So you now know how to use it, how to set it up, how to use MiniMax inside Hermes. We've tested out for different workflows. It connected to Obsidian pretty nicely, too, as well.

[00:06:13]And it's very responsive inside the chat. If you want it to be faster, I would actually recommend using the MiniMax coding plan instead. But if you want to try it out for free just to test out, then you can use something like Ollama with the cloud model, as well.

[00:06:26]The other thing that I would say about this is it's not quite at the level of Opus 4.7 or for example GPT 5.5, but at the same time you get it a lot cheaper and it's designed specifically for agentic tasks. So thanks so much for watching. If you want my full guide on MiniMax M3, how to use it inside Hermes, some of the cool stuff you can do with this, a 30-day roadmap, 100 prompts, and also my full agent operating system with all my agents plugged in to a beautiful mission control that you can just install as a zip. You can get that inside the AI Profit Board. This is an amazing community with 8,300 business owners inside here. You can ask questions. I actually answer them with daily video tutorials. You can also get access to all of my new training. So I update this daily and you can get my agent operating system as well as the Obsidian setup inside here. And additionally, if you want to know how to use Hermes for free, we've got a full tutorial out there. We do four weekly coaching calls where you can ask questions, get help and support. And this is all inside the AI Profit Board.

[00:07:23]Link in the comments description or just go to the AI Profit Board.com to get access. Thanks for watching.