NEW Chinese AI Agent MiniMax M2.1 Update!

Today we're going to be testing out Miniax 2. 1. So this is the new Chinese AI agent that's just dropped. We can try out the agent over here. We can test how it performs. I've seen actually really good things about M2. 1 and in the past I've tried Miniax and it's been a really cool AI to be fair. So we're going to test this out. See how it performs. You

can see the benefits of it here. So for example, multirogram, multi-programming language capabilities, webdev and appdev. So you can actually develop apps as well as web dev stuff. And you can see some of the benchmarks here. Let's equip yourself in the news tab. Hopefully we can see it better here.

Yeah. So you can see how it performs versus like for example like DeepSeek and other tools like that. We've got Gemini comparisons over here. This is compared to Gemini 3 Pro which is actually quite interesting. So you see for example MultiSWE Bench. If you compare these side by side, apparently Miniax is outperforming Gemini. I always take these tests with a bit of salt. You never know what's true and what's not these days. But we'll test this out and we'll see how it performs and we'll come back to it in a second. the old version two and 2. 1 on pretty much all of the benchmarks which is pretty exciting to

see as well. So when you get straight into this, this is mini max super one. As you can see, you can switch between custom, pro, and life in right now. If you use custom, you can actually set up custom instructions for your project. And auto as well is available. And then pro is like better for complex development, deep research, and powerpoint. So let's test this out. We can also use sub agents, which is quite interesting. And then additionally, we have MCPs. So we can use for example like Google maps or we can use mini maps. And we can actually add new MCPS over here as well. I've never seen it that easy to use MPCPS, which is pretty cool. And if you have any questions as you go along, like feel free to ask and we'll test this out. All right, so I'm

going to say, okay, build a website for Goldie agency SEO agency. Just going to have to sign in. I'm just going to go with the fast option for now. We can actually click on websites here, and we can actually build websites based on the templates I've got. All right, so for example, if we zoom out here, we could create like a casual cartoon style gold mining game. Let's go with that. Just see what it does. We can set up notifications if you want. Don't really want more notifications on my laptop to be honest with you. And we'll see how this performs. Re says hello. Good to see you here. You can see it's coming back to us. It's got this thinking process. It's like, I've received the task, going to help you develop it, create the game plan, and then you just have to comment in terms of what you want to continue. We've actually got a list of to-dos that we can approve if we want to keep going with this. Right? So, you can see here it's like game design, visual art, blah blah. I'm going to say just go ahead close the to-do. There we go. Right? And so you can see now it's building out the files for the website game as well. Reco China release many models projects periodically but I think they're always releasing new models. They usually focus on open source as well. That seems to be what they like to do dayto-day. Switch to dark mode. That's better isn't it? If you want to change to dark mode just go to settings and then general and you can switch between these two here. So that's beginning to code out. What we can actually do is use the same prompt and we'll just test it against Gemini Pro. So obviously on the benchmarks, if you go back here, you see on the benchmarks, for example, this is Gemini on the benchmarks. We've also got chat GBT 5. 2. So let's see how they perform and whether it's worth testing these out. So we're going to do the same thing on Gemini using Flash. In fact, no, we'll switch to Pro just to make it fair. There we go. And then we'll do the same thing on Chad GPT 5. 2, which blew me away on the recent coding test. If you haven't seen that video, check it out. Now we'll switch to canvas there. All right. So this way we can just get a feel for, okay, how good is Miniax? How does it perform versus other tools and what do we get back here? Reiko says, "Ask me, Max. " All right, then put a few. Why does China periodically release projects? If anyone's got any tests they want me to see here today as well, just let me know. So, Ji 5. 2 is coding it out. Gemini has created something pretty cool. Nice. I like this game actually. That's pretty fun. All right, let's see what Miniax does. You see here that Minia's a lot slower, right? It takes a lot longer. And we were on the lightning mode, I believe. Oh, it's on the selected pro actually. So that's probably why it's using the thinking mode inside MiniMax. This project is also open source, too. So you can run it locally. Let's have a look at some examples what it can do and we'll test this out. All right, so here's one example. Multilingual coding. We'll play this back. That's pretty nice, isn't it? Look at that. All right, let's try this out. Oh, wow. All right, this is really

interesting. So you see here we've got 3D interactive animation and it says Miniax M2 built a 3D dreamy Christmas tree based on React 3 fiber. It supports gesture interaction and complex particle animation. Let's try this out. So, I'm going to click on here. We'll allow this whilst fisting the site. Whoa. This is crazy. Look at this. Wow. I feel powerful. So, basically, if you're not sure what's going on here, this basically controlled by my hand. So, I can control the screen with my hand. As you can see here, we can use like gesture support. We can move this around, etc. How cool does that look, right? If we move the hand, it'll rotate the tree. If we do a fist, it will gather the tree. If we open the hand, it will crush the tree. Wow. Amazing. That is insane. Genuinely, I love it. Sergio says nice. Yeah, that's pretty cool. Let's keep going through now. Aentic

tool use. It's a tool use capability. This is Excel market research. Is that inside cursor? Yeah, there is. So, you can autonomously invoke Excel and Yahoo Finance to complete an end to-end task ranging from market research, data cleaning and analysis to chart generation, which is pretty cool. Let's see what else we got here. The digital employee. So, you can do that sort of stuff. End to end office automation. Yeah, I'm absolutely loving the Christmas tree though. That is one of the most impressive things I've seen this year. So, I'm going to go back to the game now. And you can see here it's taken absolutely ages to generate that. I assume it's probably rinsing through my credits as well. That's not too bad. Not too But it is slow. Let's have a look. So, Gemini created something pretty awesome. Ch 5. 2 is still coding that bad boy out. So, let's have a look here. We've got Emacs 2. 1. Wers trying

to release projects periodically. Nice little report there. Pretty useful. It's got the sources on the right hand side. It's done a lot of deep research I think on the right hand side as you can see. So the web search is really in depth and then we've actually got the files from the project as well. Kind of looks a little bit like Manis doesn't it when you're using this. Now if you haven't seen this before what I actually did previously it's quite interesting is I created a Netflix clone. This is quite a while ago on one of the old Mini Max versions. So we'll click on new task here and I think X if you switch between these you can switch between like different types of research different websites. You can actually set up schedules. You can like plan tasks for example. So, you can say at 10:00 a. m. each day, deliver a concise briefing on the most important tech and science news from the last 24 hours. Each item distilled with a single key point backed by a verified web source. And then you can actually schedule these to run each day. Right? If we ex basically got like kind of these skills, you know, I can generate videos as well. So, for example, like this, I'd love a heartwarming video of a fluffy golden retriever puppy running towards the camera in a sunlet park. Let's try that one. See what we get back from there. Ali says, "I want to build an AR rep for barber shops. Can you walk me through the process and the steps of building it? Yeah. So, what you want to do for something like that, it depends what you need, right? Depends what type of chatbot you want. If you want like a voice agent, you can build those pretty easily inside AI studio. Check out my trainings inside the AI success lab on how to do that. So, if you type in AI Studio here, use it on my best trainings on like voice agents and that sort of thing. And then also, you could use what else could you use? You could use one that I used recently was Coach Vox. Coach Vox is pretty cool for creating like chat bots as well. And you could use that on your website. So it depends where you're going to host it. But if you want a voice agent, AI studio. chatbot on a website, then you can use something like NA10 or you can use coach Fox as well. All right. So this is running the prompt now. You need to generate it. Chat GPT has taken ages to build that out. So Gemini is by far the fastest and the best even when you're using pro mode. Miniax is taking ages to generate this website. I hope it's good when it creates it. You can also get an API for this, right? So you can get APIs for this stuff. And I think if I'm not mistaken, you can actually get Miniax M2 open source on GitH. So if you go to Git and then Miniax M2, you can actually run it locally as well. Doesn't seem to be 2. 1 there, but you can run this locally. So it's open source. Then you got like mini agent as well that you can run open source to. The other way that you is it available on yes, you can get it on open router as well here. So this is Miniax. If we actually compare it, we'll run the comparisons here. So, we'll pull up Gemini 3 Pro preview and we'll compare this against Claude Opus 4. 5. Let's see how it performs. So, pretty low context window over here. Whereas Gemini has a much bigger context window, but I don't know whether they've done that for fun, but like 205k context window versus 200k on Opus. So, it does have a bigger context window than Opus. You see the activity here. So, it literally just got released last couple of days, whereas Gemini 3 has been out since November the 24th. Feels like a long time ago now. And then you can actually see the costs here. So yes, Miniax might be slower, but it's actually a lot cheaper than using Gemini 3 or using Opus as you can see. So it is a lot cheaper, too. Did it generate the answer? Yeah, we covered the answer a second ago for the channel one. Asia says, is it open source? Yes, you can get it on GitHub, I think. Yeah. So, we were looking at G4, weren't we? It doesn't. From what I can see inside Miniax inside the repositories, you can run all these different models from Miniax locally. And then you got Emacs M2 over here. This seems to be the older version of Miniax. So this is two, not 2. 1, but still you can run it locally and it's available on GitHub. So it's available. Let's have a look here. Is it available on face as well? You can't see Miniax M2 on there. I can only see M1 on Hugging Face, but you can use the inference model here. So you can test out here and that will be free. The main launcher says, "Hi, could see it. " I'm going to keep going through now. So we got the game beginning to be coded out with the command line execution. That does take a while, though. I mean, that's what it's been nearly 15 minutes. GT 5. 2 is finished. Let's select story mode. Oh, this is pretty cool. I don't know if it's playable, but it looks cool. Not as good as Gemini. I see that. All right. So, we've got this. These can run in the background as well, which is pretty cool. And then we have the video of the Golden Retriever here, too. How do we view it? There we go. It's pretty good, isn't it? Pretty good right there. Let's play that again. The quality is super nice. Like really HQ quality, isn't it? Re says, "Here's a project for you. Make rag system that supports Zim files directly. " I want to build like something more visual, if that makes sense. like something that would be interesting like a video or a game or a website like we're doing in the background. Yeah. Let's try something else now. So refresh that. Yeah. So it answered on the Chinese project. It's still building out the game with create the video and that was pretty good. So it creates videos faster than actually codes. And then if we go to new task here, let's try something else. So we're

going to say research Julian Goldie's aircraft boarding. Create a beautiful landing page for his alpha based on that research. We'll hit run on that and we'll do the same inside Gemini and Claude. Honestly, I do think it's going to be very hard for open source models to compete with Gemini and Claude right now because, you know, Gemini is just so advanced in terms of the quality of outputs are amazing. The speed at which it moves is fast. You can get free versions of Gemini as well. And then also you get a lot more functionality with Gemini than you do with Miniax, right? And so like, you know, I think a lot of people here, they want speedy outputs, they want fast stuff, etc. And you're just not you're not going to get that with Miniax, right? and it just takes too long to do. So, look at that. I mean, this is the landing page from Gemini. It's already coded out. It's already looking pretty cool. Looks great, etc. Opus is beginning to code this out as well, but it's a lot slower. So, I think the way that you would use Miniax probably like you type in what you need to and then you just come back in maybe like an hour and wait for it to load cuz there's no way you would sit here waiting to use it. Let's see how many credits we've used as well. So, we got 9,736. So, it's used about not even 10% of our credits, which is good. I try something else now. So, I'm going say create a Netflix clone with playable trailers. So, we'll click a new task. Plug that in. Hit run. See what we get. And we'll try the same with Instagram as well. Let's try a PowerPoint presentation here. See what we got. Bon says, "I hope the gold blinder game is good. Mini Max is taking too long. " I would agree. Is there a benchmark for speed when it comes to AI? That would be a really good one to know. You know what would be pretty cool is like I think a lot of these projects. So, for example, right, you look at this Christmas tree project, right? This one that's controlled by the hand. What's cool about that is, you know, that is amazing. Now, I think what you can actually do is you can search by like the last 24 hours and just see what people are creating with this stuff and that might be a faster way to preview what this is capable of, right? So, we'll do that instead. So, here's a website created by Miniax. It looks pretty nice to be fair. Um, these are indexing on Google, so you can check them out if you just search space. max. io and check out what they're doing. So, it's pretty nice right there. That's looking good. Let's try the next one. This is micro world map interactive isometric island. So we can scroll in, scroll out, etc. And you see it's created by miniax over here. Not bad. Let's try the same here. So I'm going to say to Gemini annoy Gemini will create faster. Create a micro world map interactive isometric island. We'll see what that comes up. By the way, whilst we're waiting, if you haven't already, check out the AI success lab. Link in the comments description. Completely free community. Comes with a community of 40,000 people. And you see it's very active. like people posting today and posting all sorts of cool stuff which is awesome. And then also inside the classroom here you can get all of my best trainings. So for example, if you want to get all my trainings on AI, how to use it, all the latest AI models, you can get it inside here. Every single one of these is like a course module with step-by-step implementation plans. And then additionally, if you want to get the notes from today, you get that inside the December 2025 automation section. And this runs you through like all the latest updates, including for example, Mini Max Super One, which is what we're covering today. So if you want to get the links and everything else, get that all inside the AI success lot. So here's an example. So this is from Gemini using Pro. Honestly, I would say like if I have to compare those side by side, this one looks a lot cooler than Gemini. So it does give better outputs, but maybe it just takes a lot longer. This is a kind of agency website created by Miniax. That looks super nice. So maybe the outputs are better, but it's just it takes a lot longer to generate cuz the video was really good, too. And this is a Lego builder. So we can, for example, like build stuff. Wow. Change the colors of these blocks. So, I mean, the outputs are pretty amazing. It just takes a long time to get them. All right, let's have a look at Claude Opus. So, that's generated a really nice website right there, but I think honestly for like creating it's just taking too long, isn't it, for anything? Pretty much. Yeah. So, I mean, it's a cool tool. It creates amazing outputs, but it just takes so long to use. I would not use this day-to-day in my workflows unless I was really trying to create something super complex and then I just I go off I go and do something and then this is working in the background. But overall doesn't blow my mind, you know, still great in that game. That could go on forever. And the other thing I would say here is like it is rinsing the credits now. So you see we've used about 10% of our credits whilst this runs in the background. So it takes a long time, use up a lot of credits, it's super slow. So thanks very

much for watching. If you want to get access to all my best trainings, tips, templates, workflows, etc. You can get that inside the AI profit boardroom. Inside here, you get all my best trainings. You get access to an amazing community of 1,800 builders. You can post questions, get help, get support, share your wins, etc. So we look at some of the latest wins inside the air for boardroom. So for example, Scott is on a 100 day journey. Vladimir created his own blueprint MCP. We've got Steve Light created a telegram hosting bot. Johan created his own MCP as well. So if you want to connect with awesome people, join the AR profit boardroom. Craig says very cool. And inside the classroom here, you can get access to all of my best trainings, templates, tips, workflows, etc. Inside the AI profit boardroom, we actually have full masterass on how to learn AI from scratch. So if you want to learn how to build, for example, your first AI agent in under five minutes, you can do that inside the first week here. And then it guides you through week by week exactly how to learn AI to the point where you become an expert in it. And also it comes with all my best playbooks. So for example, a lot of people ask me, how do you create the AI avatar videos on YouTube? Well, this is how I do it. This is my process. You get that right here. You can also learn how we create shorts, Twitter automations, automate emails. We're updating this all the time with new updates. So you can see for example here these are all our latest updates inside the AI profit boardroom. So feel free to get it link in the description.