MiniMax M3.1 Flash IS REALLY GOOD! INSANELY FAST & FREE! (Fully Tested)

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: MiniMax M3.1 Flash IS REALLY GOOD! INSANELY FAST & FREE! (Fully Tested) · Channel: WorldofAI · 20:58 · Watch on YouTube

Topics: Software engineering with AI, AI models, Multimodal models

Summary

The video reviews MiniMax M3.1 Flash Preview, a fast, efficient multimodal model that the presenter tests in coding and creative tasks. Demonstrations include a 3D zombie game, detailed product pages, interactive interfaces, and a Minecraft-style game, with results that sometimes compare favorably to larger models but are not always as polished. The presenter emphasizes that the model is still in preview and that some MiniMax Code subscribers could use it free during the promotion. Overall, they recommend trying it for ambitious everyday projects, particularly for its speed and value.

Key points

MiniMax describes M3.1 Flash as an efficient everyday workhorse for coding, research, analysis, and creative work; it is multimodal and still in preview.
The presenter says complete benchmark scores and deeper technical details were not yet available in the video.
The video reports a limited-time offer of free use for subscribers through October 7, with no credit burn or usage limits.
In a test, the model quickly built a Call of Duty zombie-style 3D game with rooms, barricades, weapons, and interactive features, though the presenter noted issues such as blocked doors.
The presenter demonstrates detailed landing pages and visual assets, while noting that one graphics-card page used a 2D model instead of the expected 3D one.
Comparisons with other models show competitive results in some tasks, but the presenter acknowledges that rival outputs can be more polished or detailed.
The macOS-style interface includes interactive apps and a Minecraft-style clone with block placement, textures, and crafting.
The presenter highlights speed and efficiency as the model’s main strengths, describing it as suitable for a wide range of everyday projects.

Chapters

Start	Chapter	Summary
00:00	Introducing MiniMax M3.1 Flash Preview	The presenter introduces the model, describes its intended use, and explains that its preview status means some technical details remain unknown.
01:33	Free access and testing setup	The video outlines the temporary free-use offer and explains how the presenter accesses the model through MiniMax Code.
04:10	Zombie game and early generation tests	The model builds a playable 3D zombie game, followed by demonstrations of detailed headphones and animated web pages.
09:25	Landing pages and model comparisons	The presenter reviews several web-generation examples, including an Nvidia page, and compares outputs with other models.
13:39	macOS clone and Minecraft-style game	The video explores an interactive macOS-like interface and a Minecraft-inspired game generated within it.
16:46	3D environment tests and speed comparison	The presenter shows a Valheim-themed island and village, then compares the model’s speed and output with Fable.
18:53	Final assessment and closing	The presenter sums up M3.1 Flash’s strengths, encourages viewers to try it while free access is available, and closes the video.

Quotes

It’s currently in preview, meaning that Miniax is still rapidly iterating on it. — 00:31

One of the biggest value ads of this model is its speed and efficiency. — 04:10

This is a model that keeps up surprisingly well. — 13:39

Its biggest strengths are speed, efficiency, and just being good enough to handle a huge amount of everyday work — 19:24

Concepts and entities

MiniMax M3.1 Flash Preview, MiniMax Code, MiniMax, Opus, H3, Scroll World

Full transcript (original language)

00:00 After such a hectic week, I’m finally getting some time to go over this incredible and seriously underrated model that I don’t think enough people actually know about. Allow me to introduce Miniax M3.1 Flash preview. Miniax’s latest text model, which has officially launched on Miniax Code. And considering this is a flash model, I’ve got to say the speed and quality are genuinely impressive, which is why I’m back to make a full dedicated video on it. What makes this even more interesting is that this isn’t even the

00:31 final version. It’s currently in preview, meaning that Miniax is still rapidly iterating on it. So, there’s a good chance that it gets even better from here. Miniax describes M3.1 Flash as its daily workhorse model, specifically designed for highfrequency workloads while still balancing strong capabilities with efficiency. It’s also multimodal and it’s aimed at developers, professionals, and creators across coding, research, analysis, and creative work. Miniax has specifically positioned

01:02 it for everyday development, fast, reliable, and capable of handling everything from quick bug fixes all the way to building complete features. Now, I’ll be honest, because this is still a preview model, we still don’t know everything about it. We don’t know the complete benchmark scores behind the model or the deeper technical details that the M3.1 flash consists of. I am assuming that a lot of the information will come once the model is officially released. But Miniax is giving people a

01:33 pretty crazy opportunity to actually test it right now, which is also another reason why I wanted to cover it. They originally announced a two limited time perk for miniax code from September 28th through October 7th where users could earn two times free credits from daily check-ins. And Miniax also has reset the token plan quotas when Miniax M3.1 Flash preview launched. But then they went even further. A few days ago, Miniax announced that to celebrate China’s National Day, Miniax M3.1 Flash preview

02:06 is completely free on on Miniax code for subscribers through October 7th, which is totally a steal. And according to Miniax, no credit burn and no limits. So, if you already have access to Miniax Code, this is honestly a steal because you’re basically getting an opportunity to push M3.1 Flash preview as hard as you want without worrying about burning through your credits. And that’s exactly what I wanted to do. I spent some time actually testing this model, seeing how its speed and quality hold up and real

02:37 world tests and comparing what it can produce against some of the much bigger models currently available. So, let’s get into this test and see how good this model actually is. But sorry to interrupt, guys. If you guys haven’t already, majority of you guys aren’t actually subscribed. So, it would truly mean a lot to me if you guys can go ahead and just give this video a like or even just subscribing as it truly helps me make videos for you guys on a day-to-day basis. Now, personally, like I had mentioned before, if you are to have M code, you are going to get the

03:07 best results out of this model. And this is where I have the plus plan and I use it for a lot of other things. So that is one of the main reasons why I am using it within M code. But you can see that I have it set for max reasoning with even just a 512K context window. But you can set it to 1 million so that you get better usage out of it. And since I have set it to max, I have been getting some incredible results. And it doesn’t even take that long to get these outputs when I am setting it on max when I would

03:38 assume that it would probably take multiple hours. But you can still see that it gets you some pretty incredible results. One thing to also note is that I have used 0% of my quota. This is thanks to this model being completely free for me because of the plan. This M code plan gets you access to it. So that is one of the main reasons why I wanted to highlight it. If you do not have it, it’s still perfectly fine because I believe you can use it completely for free as well within their Miniax cloud chat. And that way you can still get a

04:10 pretty good rate limit to access this model completely for free. Now, one of the biggest value ads of this model is its speed and efficiency. This is a flash model, so it’s designed to be incredibly fast, but what’s impressive is that it can still deliver some seriously impressive qualitative results. This is where I had requested it within the miniax code agent itself using the M3.1 flash preview on Max Reasoning to create an entire Call of Duty zombie style game from scratch. And

04:40 the results are incredible. Not only was it able to build everything together extremely quickly, but the quality is kind of insane. Kind of matching what I have seen from Opus if I’m being honest. Let’s actually take a look at this. The quality is actually insane, guys. This is round zero and this is where you can see the quality with the textures, the 3D environment that it was able to fully build out in 3GS. You have different rooms that you can access. You have different barricades. You even have the

05:11 features like pack-a-punch, which is in that room. I’ve actually played this a couple times cuz it was pretty fun. But this is just nuts. You even have things like the buy menu where you can buy a couple of things. You also have a light. You can even switch through different uh guns, weapons. Let’s take a look at this. Even added a dead zombie, which is kind of cool. Ow, I’m already taking health. This is our first boy. And we can see that everything seems to be

05:42 functioning as like the original game. You can see that these guys are coming from outside. They can break the actual barricade itself. Like this guy over here managed [music] to actually do that. We can take him out and then we can take out these other guys. But I’ll show you a couple other things because it even coded out multiple rooms which is nice. So there’s one more remaining. Let me go ahead and find him. Let me turn on my flashlight. I don’t know if we can escape out. This is the part that’s a little trippy. I’m outside of the building. Can’t really see much. But

06:14 let’s try to find this dude. Where are you, buddy? Can’t seem to find him. But this is what I’m talking about. Yeah. pack a bunch here. You hold E, you can purchase it. And then within the menu, you can purchase all these other things. You can uh repair the windows as well. So that way you can see where they’re at. Let’s actually see where this dude is at. You can see one error though. I don’t know why it blocks the doors off. That is one thing. But it’s just crazy to me that this model is this capable, a flash model is this capable

06:46 that it is able to output this sort of quality. And I’m not saying it is exactly like what we have seen with Opus, but it is still pretty good in terms of its overall quality. Like it got most of the functions down. You have the actual ability to purchase different products as in the different sorts of weapons. And then you have the barricades, zombies, obviously different rounds. Now, I don’t know where this dude is. So, oh, there he is.

07:16 But yeah, I’ll show you. Like you can even purchase like another weapon and that even has its own dynamic. So you have the MPK and take a look at this. You have this recoil, you have that flash animation, you even have the bullets flying out of the weapon which is kind of insane. Like a couple of years a months ago, if you were to say that this was a result from a flash model, you would be truly blown away. And that is what we’re seeing right now. And it looks like now wave two is

07:47 starting. So, let’s see how much damage we can do over here with this wave right here. Okay, I’m stuck out. Out. And I just died. If you want the best AI tools, workflows, and drops before everyone else, join my free newsletter with the link in the description below, which is completely free. Now, no lies. This is an output from the Miniax M3.1 Flash. This is just incredible cuz I have yet to see a model actually generate this sort of detail and this is

08:19 where it had created the headphones. It even created a cushion. The level of detail from this model is just insane from the actual color scheme as well as even the textures with the metal bar of the headphone. And not just that guys, it was able to obviously create a nice landing page for this as well which is incredible. And I’ll show you a couple of reference images that were tagged along with it. And this is honestly just insane that it was able to output this level of detail with the flash model.

08:51 Now, it did take a bit longer to get this output on max reasoning, which is where you will get the best outputs, but this is just something that I really never expected a flash model to output. Here you can see obviously all the details like the full landing page as well as a couple of other things that you saw from the initial page. Now, here is another solid front-end generation from this model, but this time it actually used Scroll World, which is a plug-in inside Miniax agent, and it’s designed for a scroll controlled video websites. But here, it was requested to

09:25 use the H3, which is Miniax’s multimodal video model, and it combined it with Miniax’s M3.1 Flash to handle the code, interactions, and animations to make everything feel smoother and responsive. But this is insane that it used these assets and got this specific generation outputed. Now, here is our famous Nvidia landing page prompt where we tell it to actually create a landing page for Nvidia’s latest graphics card. And this is where I got to say it is nice, but a

09:57 little lackluster. This is where usually we have most models actually generate a 3D model of the graphics card, but in this case, it had just created a 2D model. It is partially animated. It has the functional fans which is obviously great. And it has added in that reflection. So that is obviously pretty cool. You have a lot of cool effects overall with the landing page. So its front end generation quality is nice as it is with this specific model, but was looking to get a specific output for

10:28 that graphics card. But other than that, everything seems to be pretty cool with the landing page and its front-end quality is pretty nice with these generations. Now, this is a really nice landing page that it had ended up pulling putting together with different typographies, different movements, different scroll triggers, and you can see that with this output. You can see that as I’m scrolling, everything is interlocking based off of the scroll triggers. It added in different dynamic movements as well as different textures to make this more alive. And everything

11:00 seems to be polished quite well. So, that is really nice. And these are little qualitative details that you wouldn’t really expect any model to actually output. So that is overall a really solid generation from the M3.1 Flash. Now this is where things get really interesting. This is where the MiniAX M3.1 Flash on Max Reasoning was put up against the GBT6 Soul at Max Reasoning. Now the quality is just insane. Obviously, the speed is quite good with both models because the sole is pretty efficient as well. But Miniax

11:32 here was able to even get it done faster, and this is where I can arguably say that it even generated a better job than what the GPT6 did. It is obviously debatable, but both generations obviously look great, and I just am amazed that the M3.1 flash is able to get this sort of result outputed. So far, what I’ve realized is that this model goes a bit over and beyond. In this landing page, I told it to implement an SVG painting. And this is where it had animated that, adding in a

12:04 lot of texture and making this SVG quite alive. And it truly looks like an output from something like Sonnet 5.5, but this is actually coming from the Flash model. And remember guys, all of these generations are coming from the Miniax uh client, which is also completely free for me right now due to it being on the pro plan. And this is a level of detail that I didn’t really expect cuz it was able to obviously code out the SVG painting, but then implement a lot of cool features within this landing page

12:36 itself, making it more polished. This is a pakakota that was generated in voxal art. And you can see the quality here. This is where it obviously added in the atmosphere and a lot of the elements that we see state of the art models drop. But the fact that this flash variant is able to actually output something like this is really impressive. You can see the cherry blossom, the lake or like the pond and a couple of other elements that were generated with this specific output. So, I really love the output quality with

13:08 this specific model. And here is another demo and comparison. The prompt was to create an SVG illustration of Super Mario drinking a coffee with each. And this is what the GBT6 Pro had generated. Obviously, it’s a lot better. And the level of detail is definitely insane. But take a look at what Miniax M3.1 Flash had generated. This is a model that keeps up surprisingly well. Obviously, not at the same level, but it generates a pretty detailed and coherent scene, especially for a model that’s heavily focused on speed and efficiency.

13:39 I’m not saying it outright beats GBT6 Soul from this one test, but the fact that a Flash model is able to produce something like this is really competitive and really impressive. Now, it has been quite some time since we have covered the Mac OS test. And let’s actually run it with the M uh 3.1 flash cuz this is a great way for you to actually get a better understanding of how well these models have actually come in terms of its performance and its quality. This is where it looks like it

14:09 has coded out all of the individual icons in SVG. The only issue I have is like it’s not as intuitive as what we have seen with other pro state-of-the-art models, but still pretty decent for a flash variant. Aside from that, everything else seems to be interactive and complete. The top bar is functioning as it’s supposed to. Even the battery has the low power mode that you can enable, which is pretty cool. You have Wi-Fi. You have the output sound that you can play around with. You have obviously couple other things like

14:40 the notification panel that you can open up. But now, let’s actually take a look at a couple other things. You can change the background, which is nice. You have the ability to change the accents, which is actually functioning. And if you are to actually compare the results of something like M2.7, you’ll see a drastic change in quality. This is where the flash models now outputting things like what Sonnet was able to do or what Opus is able to do at the highest level. But now we can see that with the Flash variant, you have the ability to open up

15:10 Spotlight. I believe you can even do that from here as well. And all of these apps seem to be functioning as it’s supposed to, which is pretty nice. You even have a video call feature, I believe. I don’t think I can open it. Oh, yeah. You can actually call people and it has that notification. You have the actual music app that has been coded up pretty nice that plays music which is pretty cool. The podcast app. You even have the app store. If you are to purchase something or click on it, it is going to let you purchase it which is kind of cool. You have games where I

15:42 don’t know what this is. Oh, it’s tic-tac-toe. Then you have the terminal calculator and then activity monitor which is pretty cool. and then free form. But let’s actually take a look at the Minecraft clone that it has generated. This is actually pretty cool cuz it has fully added in a Minecraft clone within our Mac OS clone. And this is actually kind of insane cuz this flash model has outputed this, which is even better than what we have seen with Gemini 3.8. You have the ability to even place down uh lights, which is something

16:14 that we haven’t even seen from the other variants. You even have that break animation which is pretty cool. You can place down different blocks with different textures. You can even uh craft different things which is nice. So that is already something that we haven’t really seen from other models. So this is the level of detail that I really never expected the Flash models to actually output. For another test, it was requested to build out a vauhold themed island with palm trees, chickens, huts, and a shoreline, all inside miniax code with

16:46 no external assets. And this wasn’t a oneshot generation. But after a few iterations, it nailed the vauil style. It included everything that was asked for and even added in some extra details on its own that made the environment feel more complete. So, it is getting to a point where we’re seeing these flash models actually output this level of detail, which is something that you would truly expect from a pro variant. Here in 3GS, it was able to build out an entire 3D village terrain from a single

17:17 prompt. And honestly, the craziest part is that it only took 40 minutes to get this outputed while costing just $0. And keep in mind, this is still the preview model. If Miniax can improve it even further before the final release, imagine how impressive the Miniax 3.1 Flash is going to end up being the Flash model, guys, which is something that I really want to iterate again and again. Now, here is another comparison where the M3.1 Flash preview on max mode was put up against Fable 5.1, but on medium

17:50 reasoning. Miniax is definitely incredibly fast, finishing in around an hour and pushing over 150 tokens per second. Whereas with Fable, it took nearly two hours to get this sort of output. Now, the overall quality from Miniax wasn’t quite as strong, but remember this is a flash model and I think more interestingly the comparison would be against something like GBT 6 Luna Max and that is where you can get likely comparable results. But you can see that Fable is a bit more polished

18:21 than the M3.1 Flash. But regardless, M3.1 Flash did get you a pretty decent output while being free. It is definitely comparable to Fable cuz the results look really good. It’s just the polishing is not at the same level as the state-of-the-art models. But if we are in the future to get a pro version for Miniaax’s 3.1 series, that is truly going to blow the roof off, especially if it’s able to maintain the speed while getting obviously closer to Frontier level performance. If you like this

18:53 video and would love to support the channel, you can consider donating to my channel through the super thanks option below. Or you can consider joining our private Discord where you can access multiple subscriptions to different AI tools for free on a monthly basis, plus daily AI news and exclusive content, plus a lot more. Overall, my two cents on Miniax M3.1 Flash preview is that this is seriously a promising model, especially when you take a look at what it’s capable of doing. It’s not necessarily trying to beat every

19:24 Frontier model on raw quality. Its biggest strengths are speed, efficiency, and just being good enough to handle a huge amount of everyday work from coding, testing out different sorts of unit tests, 3D environments, games, websites, everything. It is something that consistently produced solid results within my personal test and it is pretty fast. And remember, this is only the preview. Miniax is actively iterating on it. So, it’s actually going to get better when they end up releasing the

19:55 official M3.1 flash. And I will definitely make another video on it in the future. But right now, if you have Miniax M code, this is something that you should make the most out of cuz this is a model that is free with this plan right now. So push the model, build something ambitious and see what it can actually do because you can currently experiment without worrying about burning through your normal credits. And this is something that is $0 additional usage and it is something that doesn’t have any downsides. So I will definitely

20:27 leave all the links that I use in today’s video in the description below. You can use it through the cloud or through M code which is something that you can locally install. But with that thought guys, thank you guys so much for watching. Huge props to Miniax for developing such another incredible model that is obviously even on the flash tier. But if you guys haven’t already, make sure you take a look at the world of AI school community. Take a look at the world of AI benchmark and vi coding platform, our second channel, join us on the newsletter, join us on Discord, follow me on Twitter, and lastly, make sure you guys subscribe, turn on notification bell, like this video, and

20:58 please take a look at our previous videos so that you can stay up to date with the latest AI news. But with that thought, guys, have an amazing day, spread positivity, and I’ll see you guys fairly shortly. He suffers.