Kimi K2.6: China's NEW Autonomous AI Agent is INSANE…

China just dropped KimiK 2.6. It's probably the most powerful open source AI agent ever released and we're going to be testing that today. I'll show you how to basically build and automate anything with KimiK 2.6 using these autonomous agents. You can see how it performs on benchmarks right here.

So for example, it's actually outperforming Claude Opus 4.6 on max effort, outperforming 25.4 on humanities last exam and 3.1 pro. It is performing very well. I've tested it so far. It's pretty impressive.

And this is open source as well, which is one thing to note here. Now, if we look at some of these case studies here, for example, it has long horizon coding in complex engineering tasks. What that means essentially is like it can do like more complex agentic tasks. So for example, here I actually downloaded and deployed a local AI model on a Mac, right?

Implemented it, then optimized it and it did it all autonomously, which is pretty powerful stuff. And if you want to get access to this, you can actually use it for free directly on kimi.com. So if you want to use the agent, just select agent on kimi.com and you can switch between agent, agent swarm, thinking and instant, right? So there's four different modes here.

The other thing that you can enable here is turbo speed. So there's something called turbo speed that you can enable too. And then additionally, not just that, but you can also use this in their own hosted version of open core, which you can see right here. So you can actually create, for example, your own open core hosted on the cloud.

I've actually got two different open cores launched with kimi core directly here, right? And if we go to the settings, go over to here, you can change this, you can adapt it, et cetera. Now on top of that, not just that. So just to recap your kimi, your kimi agents, your kimi agent swarms, you also have, for example, kimi claw, which is basically like a cloud hosted version of open core that you set up in one click with no expertise.

But also what we can look at here is kimi code. Now this is basically, you know, kimi 2.6 is answer to cloud code and you can launch it inside your terminals to CLI, right? And once you've installed it, basically you can use this inside your CLI and start coding whatever you want, right? So you can set up stuff locally, you can build stuff locally.

And not only that, but you can also use it inside your ID as well, right? So you can see an example right here of how you can use it inside your ID. So I'll be covering all these different techniques so you can use it and how powerful it is. And we'll be looking at some examples.

Now, if you want to see what the, I mean, the most exciting feature for me personally is the agent swarm. And what that means is basically you've got a team of agents that can automate any tasks that you give it. So for example, here, we said, go off and research the GPT 5.5 leak and create a report about it, right? And you can see it has all these different agents that's created in the team that work side by side, researching the task, and then coming up with different parts of the research tasks together, right?

And literally all we said was write a report on this. And then it goes off, it uses these models, implements this on a cloud-based computer, and then goes off and researches every step, right? Now you can see here, it's done the research, and then it deploys a team of sub-agents to work simultaneously on the same task together, right? So you've got like 12 different sub-agents working together autonomously.

As you can see right here, we've got all these different agents that are working together on the same task, which is pretty amazing right here. Let's see what we've got in the questions here. So Daniel says, are you real? I am indeed.

I'm a real person. Let's keep going now. Thanks for that question. So you can see here how it's like basically automating this task and just going off and doing stuff for us.

So I want to sort of like break this down into a useful guide in terms of how you can build and automate anything with KimiK 2.6. So let's talk about this. You've got step number one, method number one, which is Kimi agent swarms. Now these are great for like big tasks, right?

So for example, something like a big research report would be great for this, right? Now, if you want to get access to that, you can go over to kimi.com and you can start using it. Method number two is agent, right? So Kimi agent, and that's different from Kimi agent swarms because it's just one agent, not multiple agents working together.

And that means it's a bit slower to complete the task and also the task itself is going to be smaller. So let me show you an example of what that looks like. So here, for example, we used the copy from the AR profit boardroom, and then we said, create a beautiful, fun website for this, plugged it in as you can see into the chat. And then it actually creates a preview on the right hand side using the agent to generate this.

Now you can rate it at the bottom here, you can give feedback to the agent if you want to. And also you can see how it's basically created this nice little webpage for us on the right hand side. Now we can share that. So we can click on view site and we can actually share that full screen.

So you can see an example. It looks like it's pretty nice, like nice design. Let's check if the buttons actually work. Yes, they do, which is pretty good.

And that's a nice clean design that looks pretty cool. So that was created with Kimi agent. Now you can share that site once it's created, or you can set it to private so nobody can see it. And then you've also got the code here.

So you can download the whole project, all the code, and then host that on something like Netlify if you want to. But you can basically see how you can create and deploy agents in the cloud to automate whatever you want to. And for most tasks, you would just use the agent. You wouldn't use the agent swarms feature.

So if you go to agent swarm, what you can actually see is a bunch of featured case studies here. So I mean, this one is crazy. This is wild. So you can see here, it actually built our Linux.

And if we replay this task, we can see how it created its own version of Linux. And you can skip through it. You can see what's happening. You see that it's using the agent swarm to build out a full setup, and then the sub-agents it's using as well.

So these are all different agents that it's created at the bottom here. So you can actually basically create Linux like you can see right here. So we can log in as a guest, and it's created a full operating system. How crazy is that?

And that's using agent swarms. So that's one example. So these are like big tasks. These are tasks where you need an agent team to just go off and automate stuff for hours, to work autonomously for hours.

You also got here, for example, a survey. And this is a job matching system that is built. So if you have a look here, so she created like a full system for job matching people with an application tracker, everything else, right? So you can see all the files it's generated.

You can replay the chat to see how it was actually created, or just go straight to the result. And then also, if you see anything that you like, and you're like, you know, I really wish I could create my own version of that. You can click on remix here and make your own version, right? Using the original prompt, as you can see.

So it's pretty amazing. Now that's just agent swarm. You also have sheets inside here. So sheets is basically like, you can use Kimi K2.6 to build out like really useful systems inside spreadsheets, right?

So here's an example, right? It's created like a full database inside this full setup here using sheets. So if you want to like automate, say, you know, a spreadsheet or a system inside a spreadsheet, you can go with that directly. So we've got Kimi agents.

So these would be smaller tasks, plus spreadsheets, etc. plus systems, plus, you know, like websites, landing pages, tools, etc. super useful. So you got method number one, which is like great for big tasks, you know, big research report or a full operating system.

And then you got method number two, which is like Kimi agents, smaller tasks, spreadsheets, systems, websites, landing pages, tools, stuff like that. Now, you've also got like kind of the chat directly. And you got two options, thinking and normal. So if we go back to Kimi here, you can just use instant or thinking, right?

And these are basically like just, you know, it's a quick chat. It's kind of like chat GPT, not really agentic, but you can still use it. And so you can say, okay, what time is it? I don't know why you would ask you what time it is.

But basically, that just gives you an example of how quickly it can respond if you're using the instant mode. So if you just need like quick answers, then you go with that. Now over here as well, you've also got like slides, you got websites, you got docs. So these are kind of like different ways that you can implement this into your actual work, right?

So for example, you can use Kimi to create a slideshow, like you can see here, you could use Kimi to create a full website, right? But you can also remix other people's websites, and then just check out those and then create your own based on that. It also got deep research, right? So like super deep research tasks.

You see an example here. So it's really deep, you know, it's getting all the numbers, it's doing research across like multiple different studies, and then piecing it all together into like a nicely designed and organized page here. The other cool thing about this was like it comes with a nice contents as well. So it's like an amazing interactive research report that's super useful.

What else do we have? Yes, we've got slides, websites, docs, deep research sheets, and then, you know, your Kimi code. But let's before we come on to that, let's talk about Kimi Claw, right? So with Kimi Claw, basically, this is a cloud hosted version of OpenClaw, right?

So we can like schedule tasks inside here, it can work autonomously, 24 seven, we can access it from our phone, etc. So if we say, okay, schedule, researching AI news daily, we can schedule tasks. Now this will run 24 seven, with or without us opening up Kimi, right? And so you can set up Kimi Claw in the cloud, you can always add a new claw as well.

And then you can switch between like on your phone, on your computer, or on your cloud server. So if you do want to run this locally, then you just use one of these two options, right? You don't have to run it on the cloud based version. But you know, if you want the easiest non technical way to set up OpenClaw with Kimi K2.6, this is one of the best options, right?

So you can select on cloud server here. And that's method number four. So just to recap on that, basically, you use Kimi Claw for scheduled tasks in the cloud, you know, this is autonomous, running 24 seven, it's the best option for managing AI agents from your phone, right? And it can be local setup if you want, or just one click setup, the cloud hosted.

And again, some people say, well, what's the, you know, what's the benefit? Scheduled tasks, right? What's the other benefit? Autonomous running 24 seven, no need to keep it open.

And it's the best option for managing agents remotely from your phone. And then the other benefit as well, of course, is like it's non technical, right? 99% people don't know how to set up OpenClaw, Kimi Claw solves that. By the way, if anyone's watching and you have questions along the way, like feel free to ask, that's what I'm here for.

So that's four different options. And then you know, the final and fifth option is Kimi Code, right? So let me show you how to set this up. So Kimi Code is over here, I'm going to run this inside a terminal.

This is basically Kimi's answer to Claude Code. The difference is like it's cheaper, and also you get more usage, right? When I've tested them out side by side, I would honestly say like, Kimi Code is not bad, but it's it's still not as good as Claude Code, right? Like Claude Code is the goat as far as I'm concerned.

So from here, once you've installed this, you just copy and paste this command from the terminal. And I'll just put this here. Bear in mind, once you've installed this, you just need to be on the coding plan to use this with Kimi Code. And then once you've set it up, you just run Kimi, right?

So if we type in Kimi there, it will start creating the agent, right? And now we can speak to it directly. Now, also, you can see here that we need to update it because it's just come out. So I'm just going to do that now.

Alright, so let's run it now. And then boom, we can start using Kimi CLI, right? So if we say okay, or let's try this, we'll say okay, create a beautiful productivity tool as a one page HTML, like a fun pomodoro timer. Alright, that'll be able to code directly for you.

And then the final method as well is like you've got the IDE, right? I would sort of plug them in directly there. So you got terminal CLI, or you've got the IDE. So you can install this in virtual studio code.

And these are the plans, by the way, in terms of I'm on this plan right here, because that gives you access to Kimi CLI as well with the setup. And I've come nowhere near to hitting the limits when I've been using Kimi. So it is pretty powerful stuff. So let's talk about this, you know, Kimi K 2.6, how does this work?

How can you use etc? So, you know, what even is Kimi K 2.6? It's basically an AI, right? I think we like chat, Chibita, or Claude.

But it's special because it's designed for agentic tasks, right? Now, this comes directly from China. And it's also open source as well. Why does it matter for you?

Because it basically lets you work more autonomously, right? You know, research from McKinsey actually shows AI agents can save 60 to 70% of your daily time on tasks. That means, for example, if you work eight hours a day, you can get the same work done in two to three hours, right? Or you get 3x done in the same time.

That's the power of Kimi K 2.6. That's what we're looking at right here, right? And so there's five different methods. So your agent swarms are using this great for big tasks, you got Kimi agent, single agent, which I've shown you today, you know, field and spreadsheet, website, landing page, calculator, etc.

These are great. You've got the chat and the thinking, which is a simple way. It's just like using chat Chibita. You know, if you need quick questions, or you're brainstorming or that sort of thing, or you need a fast answer, and then you've got Kimi Claude, right?

So basically, this lets you run AI agents in the cloud. What does in the cloud mean? It means the agent runs a computer somewhere far away, right? You don't need to keep your laptop open.

You don't need to keep it running. It just works 24 seven on the side. Now, why does this matter? Because you can schedule tasks, you can say every morning at 6am, go off and do this.

And you can manage everything from your phone, or you can set up one click. That's basically how it works. And that's, that's useful. Kimi Claude is useful for any tasks you want running in the background, right with daily reports, scheduled tasks, etc.

And then you've got Kimi code, which is similar to Claude code. And that's basically it. Now, if you want a full guide on setting this up, I'll put it inside the AI Profitable Boarding. So just go to the classroom here, new advanced AI tutorials, where the page and you can see here, it's heavily, heavily focused on on AI agents, right?

If you look at all the benchmarks, all the tasks, etc, is fully focused on AI agents, which is what we want. Also got long horizon coding, right? So it can just run autonomously for hours. That's the difference here is like it can just run on its own for hours and get stuff done, which is great.

And they've got like 11 different case studies and examples of how that's worked with different companies have implemented it. Now, there's also something called coding driven design and see how it performs on design bench here. Some examples of like the styles it's done, right? So these designs are super nice, right?

These are all examples of how it uses the coding style for creating websites. And also, this is something interesting as well as it can actually generate videos and images. I've not even come on to that. So let's have a look at that in a second.

All right. So let's say, for example, you're in OpenCLAW, or you know, you're in KimiCLAW or whatever, you can actually create images. So if we go to a new chat here, we'll just do instant thinking, you know, let's do instant. And we say, okay, create an image of a cat eating a banana.

And you can either do it this way, or you can use something called Kimi designs, right? So you can actually use this and it's kind of like a design agent for creating better design. So if you're creating something more complex, like a website, you would go with the agent, right? But if it's just like a quick image, then it will give you options on how to create it.

And that's basically how you can use it. And then you can also go over to the chat here. And you see how you got these skills as well. This is interesting, too.

So you can actually set up these skills and add new skills with Kimi, which means like, basically, these are like individual experts that you can train up Kimi with, right? So for example, let's say you wanted to make it really good with SEO, where you could create a new skill for SEO with Kimi. And then basically, every time you create a blog post, you can use that skill, generate content and also publish it and that sort of thing. So you can see how that works right here.

And then we can say, okay, create a skill for me using skill creator, his requirements. And then let's say, for example, you were doing this for SEO, where you could actually take a prompt like so, right? Give it your ideas on how you want to write the content, which we've got over here on this prompt. Then if you go back to the skill creator over here, Winnipeg saying, here's my requirements, and it will create a new SEO skill, right using Kimi computer, which is a computer in the cloud that's just operating and you can preview it right there.

The only thing I mean, I'll be 100% honest, like the only thing I don't like here is how long it takes, right? Look how long that takes to load up. It does take a while to start these virtual computers and create the skills. But at the same time, you can just run it in the background.

Now, also, if we look at some of the other stuff here, so it says, this is for proactive agents, right? So KimiK 2.6 demonstrates strong performance in autonomous proactive agents, such as OpenGL and Hermes, which operate across multiple applications with continuous 24 seven implementation. And so you can use these, these are designed to work really well with AI agents. And look at this, like it worked five days autonomously on this particular workflow, which is pretty crazy, right?

So KimiK 2.6, operated autonomously for five days, that is absolutely insane. And you can see how it performs on KimiCore bench, which is basically like a benchmark for measuring how powerful is with AI agents like OpenGL. And then we also have the agent swarms examples, right? So you can see examples of how this works right here.

So it can like match 100 jobs to 100 resumes, find 30 stores about websites and generate landing pages, you know, and it's just using all this agent team functionality to go off and do that, right, which is pretty crazy. So a super powerful tool, really good option for running with AI agents, easy to use, easy and intuitive to set up, open source, and you can get access for free as well if you go to kimi.com, right, use it inside the chat. So if you want to get my full guide on exactly how to use KimiK 2.6, you can get that inside the AI Profitable Boarding Pair. As you can see, it comes with a 30-day roadmap and a step-by-step SOP on exactly how to use this stuff.

We also got a full six-hour course on OpenCLR and another two-hour course on Hermes right here. And we had new daily trainings for advanced new tutorials right there. Now this is my community for helping you save time and grow and scale with AI automation. So you can ask questions inside the community, get my weekly updates, stay ahead of everything.

You can jump on weekly coaching calls where we go deep on AI agents like Kimi, OpenCLR, etc. And you can ask questions, you know, ask about your setup, share your screen, etc. Inside the map, you can meet people inside your local city who are doing similar things to you. So if you want to meet people using AI agents inside your city, you can check out here.

And then inside the classroom, you get new trainings all the time and all my best courses, as you can see. Plus, you can connect with me personally. So that's all inside the AI Profitable Boarding. Link in the comments description or just go to the AI Profitable Boarding.com to get access.

Thanks for watching.