In this video,

I'm going to be vibe coding with MiniMax M2.1,

which is an AI model that a ton of people are using in

their vibe coding workflows and that is ranking very high

on vibe coding and coding leaderboards.

So with that being said,

before we dive into actually vibe coding with this model,

where I'll share a couple important things that you should

know about the model, I do want to ask that you like,

subscribe, and join the Bridgemind Discord community.

Also,

note that we do have a light goal of 150 likes on this

video.

So if you haven't already, make sure you like the video.

And with that being said,

let's get right into testing out this model and vibe coding

with Mini Max M2.1.

So first things first, LM Arena, you can see here,

it's ranked seventh on Ella Marina.

I think that LM Arena is good to look at for leaderboards

just because, you know, people,

how this works is people use it and then they vote on which

models are better.

So you can see it has 6,387 votes, which is pretty good.

It has a score of 1422.

So it's going a little bit under GLN 4.7.

But going back over to OpenRouter,

important things to know,

this model was launched on the 23rd of December.

It has 196,000 contacts,

27 cents per million on the input and $1.12 per million on

the output,

which from a cost perspective does put it under GLN 4.7,

which is another open source model that a ton of people are

using.

So if we go over to artificial analysis,

which is a great view,

not only that LM Arena has where it has the voting,

but artificial analysis does a little bit more in-depth

analysis of these models.

So here you can see how it's performing when compared to

some of the Frontier models.

So when you look at this, you can see, first of all,

Mini Max M2 here to Minimax M2.1.

So they were able to improve the model, obviously,

but it is performing worse than GLN 4.7,

and it's performing way worse than some of the Frontier

models like GBT 5.2, Claude Opus 4.5, and Gemini 3 Pro.

Now, when you go down here,

I do want to draw your attention to other important

leaderboards.

Some of the most important ones being, where is it?

It's the Humanities Last Exam.

You can see how it does there, 22%.

You can see coding here, 41%.

So you can see that it performs well,

especially when you take into account the price,

when you start thinking about not only, okay,

how does the model perform, but what is its price?

I know that there are a lot of people out there that price

is one of the most important things when they're

considering using a particular model.

So when it comes to Minimax M 2.1,

it is by far the most cost-efficient model that you're

going to be getting because it performs decent and then

it's super, super cheap.

Like, hey,

if you want to look at the difference between Opus 4.5 and,

you know, Minimax and 2.1, it's substantial, right?

I mean, look at this, Opus 4.5,

$5 per million on the input,

$25 per million on the output versus minimax,

which is basically, I don't even know what that is, what,

a 20th of the price?

So it's definitely cost efficient.

You can also use this model for free using open code.

So we are going to be vibe coding using Bridge Code today.

So as you guys know,

Bridge Code is a fork of Open Code where we're building

additional features on top of Open Code and we're staying

up to date with upstream updates from Open Code.

But Bridge Code is going to be a way where we can integrate

with the Bridge MCP.

We can integrate with all of these technologies that we're

building at BridgeMind for our vibe coding suite.

But I'm going to launch Open Code because you can actually

use this for free inside of Open Code.

So check this out.

So Minimax M2.1, you can use this model.

There is some rate limiting,

but you can literally use this model for free on Open Code.

So if you are wanting to try this out for free,

you can use it on Open Code.

Now, for today,

we are going to be using it inside of Bridge Code.

And you can see that this is actually the BridgeMind IDE

that we're using now.

We're probably going to come up with a different name for

it, but this is one of the projects that we are working on.

In Bridge Code,

it's cool because you have a couple different modes.

You have diagram mode, bridge MCP mode, ship mode,

docs mode, vibe mode,

and then you have the native open code build and plan

modes.

But we are staying up to date with upstream updates from

open code,

and then we're building around BridgeMind and the

BridgeMind sphere of vibe coding products.

So with that being said,

let's go to models and let's actually integrate with

MiniMax M2.1.

So inside here,

we're actually just going to search up OpenRouter.

All right, so OpenRouter has a lot.

So let's go back here and let's just do Mini Max M2.1.

Here it is right here.

And we'll also change to it here.

Models.

And then we'll change to Mini Max M2.1 right here.

And then same thing, models, minimax M2.1.

Okay, so we're in Minimax M2.1.

And what I want to do today,

what we're going to be vibe coding with,

is we're going to do a couple different things.

First of all,

we're going to be doing some UI updates to the Bridgemind

site.

So one thing that I want to draw your attention to is the

Bridgemind Vibeathon.

So as you guys know,

this is something that we've been planning on rescheduling,

and we are going to be making some updates to this page.

So here, this was initially going to be,

it's a Vibe-a-thon where you build an open source tool that

makes Vibecoders unstoppable, $1,000

in Bitcoin to the builder who ships the most impactful

project.

So as you guys know, this is going to be now rescheduled,

and we're actually going to be making those updates using

MiniMax M2.1.

So I'm going to use Whisper Flow here and I'm going to give

it my prompt.

But one thing, I mean, we can just dive into it later,

but let's just give it a prompt.

I want you to evaluate the website and I want you to take

particular note of the Bridgemind Vibathon.

And I want you to update some of the content related to the

Vibathon, particularly the dates of the Vibathon.

The Vibathon will stem from February 1st,

2026 to February 14th, 2026,

and there will be three winners competing for $5,000

in prizes.

The first winner will get $2,500.

The second winner will get $1,500.

And the third place winner will get $1,000,

totaling $5,000.

So I need you to review the banner on the website and

update that with the new amount of prizes.

I need you to update that with the new amount of winners.

And I need you to update the dates for the Vibe-a-thon and

some of the content so that it's updated in this way.

So here's this prompt.

I just pasted the text in and let's send it off.

So here's the prompt.

We're going to be making a couple of these updates.

And the next thing that we're going to do,

so while this is working,

you can see here is the process of its thinking.

You can see the amount of context used and the price here

that we're getting billed through open router.

So so far, just a penny.

So again, you know,

this model is going to be very inexpensive.

You can see it was able to read the correct pages and find

the correct pages and now it's updating them.

It's understanding the current state.

So that looks good.

But let's get started on another bridge code window.

I am having an issue.

So as you guys just saw with me using Whisper Flow,

we are working on a new product called Bridge Chat,

which is a voice to text tool that is going to be in the

BridgeMind ecosystem of vibe coding products.

So one thing I want to draw your attention to is right now

we have all of this running here.

And if you look here,

I'm having an issue just running this.

So I do want to figure out what's causing this issue.

Yesterday on stream, we did actually like,

we moved a couple things around because we decided that we

were actually going to put the Bridge Chat inside of the

Bridgemind ecosystem rather than having Bridge Chat be like

a separate brand or product in itself.

It's actually going to be included in the Bridgemind

subscription.

And remember,

we are planning on launching this suite of products in

February of 2026.

So with that being said, let's now at Bridge Chat Desktop.

Okay, this is the application.

And I'm going to paste in the error that I'm getting here

and I'm going to just submit it.

And we are in vibe mode right now.

Okay.

So this vibe mode is going to be able to,

this is a new mode for Bridge Code.

And that's one of the reasons that we're building a fork of

Bridge, of Open Code.

And we're building Bridge Code is because Bridge Code will

be able to interact with the BridgeMind suite of products.

So one of those things would be, for example,

just like the BridgeMind MCP, right?

The Bridge MCP,

where you can actually interact with this Bridge MCP

natively.

And rather than having to, for example, like prompt, hey,

I want you to use the Bridge MCP, you're just able to,

you know,

toggle it in Bridge MCP mode so that it natively works with

the MCP and it's able to manage those tasks respectively.

So let's check this out.

What is it doing?

It looks like it's checking the Tori config schema.

So this should be able to fix it and then we'll move

forward.

It looks like it's just an issue with the schema file.

So how does this look?

Okay, this looks fine.

All right, so it's changing that there.

And now it says try running it again.

So let's see if we can actually run this again.

Let's see if Bridge Code was able to fix it in one shot

with Minimax M2.1.

Let's go.

And it looks like it.

It worked.

It works.

Okay, perfect.

So Bridge Code with Minimax M2.1.

Okay, it fixed that issue, but now there's another issue.

So let's now copy this issue and fix it.

But it was able to fix that initial issue.

So let's now paste in again this new issue.

I am now getting this error fix this.

And you can see that this agent on the left is still

working and we'll be able to pull up what it's doing.

But it looks here like this is updating the Vibathon page

with the updated information about the Vibathon.

Really excited to do that Vibathon.

Votes on submission and winners announced live on stream.

That's great.

Okay, perfect.

Okay, so it says that, let's see here.

Com bridge chat.

Okay, perfect.

That's going to be able to work through this.

And okay, remove zero files.

The card didn't find anything.

The ROS tool chain.

Okay.

So it's going to kind of reset some of these things.

And I think that that will be able to fix it.

But I really like using Vibe mode, but Minimax M2.1.

Let's actually see how it's doing on the website updates.

Okay, so here you can see Bridgemind Vibe-a-thon,

February 1 to the 14th of 2026, $5,000 prize pool,

three winners.

So it did everything perfectly.

And I would actually say that it totally maintained the

integrity of the styling of the page as well,

which is very interesting.

So you can see that not only was it able to update it,

but it was able to maintain the integrity.

And the big question that a lot of people will probably ask

is, hey, how does it look?

What's the responsiveness?

And the responsiveness is pretty good as well.

So look at that.

Really good responsiveness, even on mobile.

Now, I am going to look up another issue.

So I had an issue reported on the Bridgemind website that

for some reason,

people were having issues scrolling on different browsers.

So what I want to do now is I actually want to make it so

that we're going to be, and it looks like this.

Okay, so this is on my other screen right now,

but I do have the option to sign in for Bridge Chat.

So I do see Bridge Chat running here.

There's a sign in button for it,

but Bridge Chat is now running.

Can I pull it up?

Let me see here.

Let me see if I can find Bridge Chat.

We're going to have to update a few things with that.

But let's now launch this another agent,

and we're going to switch to plan mode for this with Minmax

M2.1.

I want you to review the Bridgemind UI,

and I want you to identify why some users on different

browsers or different computers are having issues with

scrolling.

People have reported that on some pages they are unable to

scroll or there is no scroll bar.

I need you to review the application and figure out why

exactly that is happening and create a plan with updates to

fix this issue.

You need to diagnose and pinpoint exactly why this is

happening.

Okay,

so we're going to submit this and that's going to go in

right there.

And this, we're going to have to cancel out of this.

It looks like this is downloading and building.

Let's collapse it.

One thing that I need you to review.

Okay, let's actually launch another.

So all I have to do is I'm going to keep it here.

I'm going to launch a vertical pane.

And you can see like, hey,

this issue here was very cheap to update.

I mean, if I used like Gemini 3 Pro for that,

it probably would have cost like, you know,

a dollar for this update,

but it only cost eight cents for Minimax M2.1.

So I will say, I mean,

it got the job done perfectly and it was able to do it in a

way where it didn't cost me a bunch of money.

So, you know, it's not the best model.

It's not the smartest model, but I will say that, you know,

it got the job done very inexpensively.

So let's change over to build mode here and let's add

bridge chat.

And there's just one thing I want to do.

Currently,

there is a sign in button that is required and I want you

to make it so that this is not required for now.

I don't want to worry about authentication inside of

BridgeChart right now.

I want any user to be able to use it, review this logic,

and I want you to update it so that we remove the

requirement for authenticating for now.

Okay, so there's that.

Let's just build this out.

This one here, remove the invalid protocols.

Okay, perfect.

The build is now compiling successfully for Bridge Chat.

So you guys can see that, hey,

it nailed our front-end updates and now it's working on

some of the, you know, this Tori app.

This Electron stuff is good as well.

This over here, it's investigating and creating a plan.

It's going to be able to review everything and then create

a structured plan to figure out this error and then

diagnose it and then fix it.

Here, it's saying that I can see the issue.

It's going to fix that gate.

So I do have another thing that I want to say.

I want you to review the positioning and the styling of the

bridge chat icon.

So right now it's positioned above the Mac OS toolbar,

and I need you to make it so that it is, okay,

it's actually the same thing with Whisper Flow.

So let's try this out.

We're going to launch it again.

I'm actually not going to submit this.

I'm going to wait for this one to finish because I actually

don't want the auth wrapper there yet because that's a

little bit more of a complex issue.

Maybe we can actually create a structured plan for that

though, because, you know, for example, hey,

even when I'm using the Bridgemind workspace here,

which you guys are seeing and I'm using right now to be

able to run all these agents, I'm using that right now.

And that is off-gated, right?

I have to have a BridgeMind account.

I have to be signed in.

But we don't want to worry about that yet with Bridge Chat.

I want to use a little bit of a smarter model potentially.

We could try that out in this video and see if we could

actually build that out with Minimax M2.1 and Bridge Code,

but we'll see.

I want to wait for this one to finish and then we'll

relaunch and try again.

So let's actually open up a new workspace.

And I'm just going to name this one MPM.

And I'm going to put this in red so that I know, okay,

this is going to be the workspace where I'm managing all of

my different terminals.

So let's CD into my desktop and then let's CD into

BridgeMind and then let's CD into Bridge Chat desktop.

And now let's run npm run Tori dev and let's see what

happens now when I run this.

Okay, okay, okay, perfect.

Okay, so it's now running.

It's going up.

All the models were loaded successfully.

And I do now see the bridge chat.

Okay, perfect.

Okay.

So I do have bridge chat running.

I do want to fix one thing though.

So let's pull up this workspace again.

Okay,

one thing that I want you to take a look at is right now

with Bridge Chat, there is this icon.

And with the icon at the bottom,

I'm able to talk into it and there's an effect.

But I want you to make it so that this icon is positioned

on, you know, the screen that the user is focused on.

Right now,

it's hard-coded to be positioned above the toolbar on a

Mac.

I want you to fix this so that it's updated to make it so

that it will be positioned on whatever screen that the user

is focused on.

So you guys can see like the issue that I'm having is that

you guys can't actually see the bridge chat.

You guys can't see this right now, right?

But we want to make it so that, and here's one thing,

I wonder if, does MiniMax M2.1, does it take in images?

I'm actually not sure.

So let's actually turn this into plan mode here and we'll

kind of create a structured plan.

But here with Bridge Chat, I now have the ability to, okay,

so here is Bridge Chat,

and you guys can see that I've spoken 3,650

total words to it.

Speaking time, 23 minutes, 170 sessions, words per minute,

155.

Was this one able to figure out what the issue was?

So you can see that it did launch.

Okay, so it says control X right to view sub-agents.

So let's see.

Can we, okay, wow.

So here are all the sub-agents that it launched when

planning out this task.

So you can see that this is still working here and we'll

let this finish working.

But this actually is doing a pretty good job.

I mean, it nailed this.

It was able to remove the off because I can now use Bridge

Chat.

You guys can't see the icon, but for example,

this is Bridge Chat.

This is a quick test of Bridge Chat.

So there it is.

This is a quick test of Bridge Chat.

That is Bridge Chat doing that.

So that is going to be a product in our Vibe coding

workflow.

All I have to do is I can set it up where the hotkeys are

anything that I want.

So if I just, you know, let me just force quit,

can I force quit Whisper Flow here?

Let's hold on.

Let's pull up Whisper Flow and then let's force quit

Whisper Flow and we'll go on from there.

So let's pull up Whisper Flow and let's just force quit

Whisper Flow and then we'll only use Bridge Chat.

Okay, so you can see that we're launching BridgeMind,

which is an Electron app and then Bridge Chat,

which is the Tori app.

Let's force quit Whisper Flow.

So you can see that that just went away there.

And now we're only going to use Bridge Chat for the rest of

this video.

So this is going out with this plan.

And again,

we can do the control X right to launch any sub-agents.

But I will say that what I'm seeing right now from Minimax

is it is getting the job done for a very,

in a very inexpensive way.

Okay.

It's working.

It's thinking a lot.

It's getting the job done.

Now,

you could make the argument of the fact that the speed isn't

great.

Like if we go back to Open Router, what is the speed?

I mean, the speed on OpenRouter is pretty good.

Atlas Cloud, it's not that good at 25 tokens per second.

But for the most part,

the speed is pretty good at 72 tokens per second.

Mini Max, it's 57.

How does that compare to GLM 4.7?

Well, Opus 4.5 is like 36 right now.

So Opus 4.5 isn't that fast anymore.

I don't know.

It's kind of like been deprecated.

Maybe there's just so much demand.

But you can see that, okay,

GLM 4.7 is a little bit slower than Mini Max M2.1.

So Minimax M2.1 is, you know, on average faster,

at least right now, than GLM 4.7.

I wonder if we can see this inside of the speed here and

the speed.

Is there a speed index that we can get to?

I don't know how to get to that speed index.

So I do think that it is faster and it's getting the job

done.

You can see that it's working through all of these and you

can see that it actually is finding some of these scrolling

issues and this one is also planning out.

So we'll let it let it work for a while,

but I will say that when you're using bridge code and

you're in plan mode, it does plan for a while,

which is a good thing.

Let's go to this one and let's use bridge chat and give it

a prompt.

I want you to now run npm run build and fix any build

errors.

So we're going to run npm run build.

We're going to fix any build errors and then we're going to

deploy that to production so that that is updated.

There'll be some more information about the Vibathon here

in the next live stream.

So that's going to go up.

It looks like the build is going out successfully.

And here is this bridge chat here.

This bridge code agent is done.

Now, what is this here?

So you can see that this is what it found.

And can we now, how do I?

Okay, cool.

All right.

We just tagged that chat.

That's all right.

All right, bridge code.

Is there a way for me to resume past conversations?

Resume list sessions.

Okay, so which one was it?

It was this one here.

So we'll open that session back up.

And let's actually switch out of plan mode and let's go to

ship mode.

And I'm going to use bridge chat.

I now want you to update the code to fix this scrolling

error.

Okay, so let's update the code to fix the scrolling error.

What's this one saying?

So this one is saying update the main to use this.

Okay.

Okay.

Alright, mouse position.

Okay, make the updates with best practices.

Alright,

so we're going to change this one to ship mode as well.

And this is going to be able to build that out.

So the root issue here,

the scrolling problem is caused by a CSS stacking context

conflict between multiple overflow X hidden declarations.

So it was able to find the root issue here.

This one is going here and fixing any build errors.

Well, that looks good.

Okay.

All right.

All right, so that'll be fixed.

This one was able to move the off.

This one here is going to try and fix the mouse

positioning.

So the issue here that this one's trying to fix is,

you know, this bridge chat, it's fixed under the toolbar.

And really what it should be is it should be fixed in

whatever screen I'm focused on.

So we just need to update some of the code to be able to do

that.

So we'll see what that,

if that's able to get it in one shot.

Now here, let's see on a flex container with Flexer.

Okay.

All right.

This looks good.

This one's working well.

This one's in build mode.

This one's in ship mode.

This one's in ship mode.

This one.

Okay, let's use...

I want you to review the project in complete depth,

and I want you to take particular note of how we are going

to handle auth.

Would you look at the following projects?

So the reason that we kind of had to rethink this approach

is because there's some code in the,

or actually not the web app, it's the, okay,

and I actually just lost that, lost that prompt.

Okay,

so this one is now updating and I can't even see the screen.

Okay, is it still working though?

Let's see.

This is a quick test.

Okay,

so this one actually did make some adverse updates here

that we'll need to fix.

But all in all,

like one thing I do want to draw your attention to is that,

okay, look at the cost that's being spent for this, right?

12 cents, 4 cents, 3 cents, 1 cent, right?

And you can even use this for free inside of OpenCode if

you don't want to spend anything.

But I think that, you know,

MiniMax M2.1 will get the job done.

It won't get the job done as efficiently or as correctly as

some of the Frontier models.

So what I will say is that, hey,

I think that for tasks that may be related to simple UI,

you're going to save a ton of money by using a model like

Minimax M2.1, where, you know,

Cloud Opus 4.5 may be overkill for that task, right?

So why spend a dollar using Opus 4.5 when you could spend

10 cents using Minimax M2.1?

That difference may even be more.

But if we now go down here,

what I do want to draw your attention to as well,

which is important, is the hallucination rate.

Okay, so this is a hallucination rate index.

Let's go over here.

And what you can see is that Opus 4.5,

one of the reasons that this model is so good at vibe

coding and at coding in general is that it does have the

lowest hallucination rate.

So you can see, check this out.

The hallucination rate for Minimax M2.1 is actually pretty

low.

It performs better than GBT 5.2 and better than Gemini 3

Pro.

And if you do go to this one, it is not that great.

It's minus 30,

but it does perform much better than Mini Max M1, M2,

I mean.

And then for the accuracy, you can see, ah,

this one is not great.

So there's kind of some conflicting views there, but here,

higher is better.

So this performs not super great.

But again, I think that for this model,

I am going to give it the Bridgemind stamp of approval

because, hey,

just the cost benefit of this model is so significant that,

hey, if you go back to our ViBathon page,

it did everything that I asked perfectly.

So you may have to be a little bit more specific in your

prompts, but at the end of the day,

you're probably going to save quite a lot of money using

MiniMax M2.1 in your workflow because, I mean,

look at this model.

It performs better than GPT-5.

So it's like, hey,

you have a model that is literally like 27 cents per

million on the input and like close to a dollar on the

output, super cheap.

That's performing better than the Frontier model that we

had less than six months ago.

So you can see how fast this field is changing.

You can see how fast it's evolving.

I'm going to approve this model.

I think that it's getting the job done.

I'm going to continue these updates off this video,

but definitely was able to make the correct updates and get

the stuff done that we needed to.

And I am going to be using this model in my workflow for

like some simpler tasks that I don't need Opus 4.5 for or I

don't need any Frontier model for because it's going to

save you a ton of money.

So with that being said, guys,

I will see you guys in the future.

Make sure you like and subscribe if you haven't already.

And I will see you guys on stream later this evening.

See ya.