MIMO

Your waifu is finally getting delivered.

Xiaomi releases their latest and best model, Mimo 2.6. XAI also releases their best model, Grok 4.7. OpenAI drops their latest model, GPT-6 Sol and Luna.

And then Anthropic also drops their best and latest model, Opus 5.5. And then shortly afterwards, Google also drops their latest and best model.

Oh, nevermind.

If you're interested in reading further, I'll link to this main page in the description below.

Now, last week I showed you

this dashboard where Xiaomi shared their reinforcement learning process for their upcoming model MIMO 2.6 in real time.

So you can see the training runs and the progress of the model over here.

You can put it on your desktop computer and run it at home to generate images for free, and it's made by Alibaba.

Xiaomi put out Mimo on September 22nd, two days ago.

Mimo Pro performs on par with Claude's Opus 5, which is, like, this insane model that everyone thought was gonna change the world and break the internet, and GPT-56 Soul across most benchmarks.

It's a 309 billion parameter model.

It is entirely open weight.

No, just move the, move the workloads where they belong and save the money.

Yeah.

And so Sax, um, I made a chart here before the show here, but here's your, um, uh, sweet spot for the last 100 days of models, which, um, Triber, you did a good job of teeing up, and what you'll see [laughs] is like, yes, the Claudes and the Astras up in that right-hand corner cost a lot, and they deliver a lot, but then you start going down, and you start to see Muse and GLM and Kimmy, and then on the left, you start seeing GLM from ZAI and Mimo, and you're starting to see, well, the Chinese ones here-

The Anthropic billing bomb, obviously.

Then Xiaomi's MiMo model one-shots an entire playable game for thirteen cents.

All right, and the closer today is wild.

A name you'll actually recognize joining a lab most builders sleep on.

So Xiaomi.

Here's the thing that got me.

Their new model, MiMo version two point six Pro, just posted forty-six points on the Artificial Analysis Intelligence Index.

GPT five point six Sol sits at forty-seven.

So, so you think like, uh...

So there was, there was another, um, uh, ultra fast model that also got released yesterday, uh, Mimo.

Um, and I've, I've got a Mimo account, but I haven't, I haven't tried to use this yet.

Um, so these guys, um, have a new architecture that's super fast.

But what I find really hilarious about this, like the Chinese labs are like trying to differentiate themselves, right? So they have this model that's super fast and quite smart.

The GPT fast mode is like 1.5X faster.

These [laughs] fucking Chinese guys are like, "1.5 X, man? No, no, no.

20X." And so the cost of output tokens on this model, uh, 2.6 or whatever it is, Mimo, Mimo 2.6, um, is 80 cents, right? And so what they do is they basically throw 20 times more compute at it, so it runs 20 times faster.

MIMO V2.6 from Xiaomi just dropped.

And this is a powerful new model.

It's actually frontier level.

It's open source.

And not only that, but also you can use it for free and you can use it inside a Hermes agent.

So Xiaomi just dropped MIMO V2.6.

It is the strongest open source model ever released.

It has two different versions, a Trillium Parameter Pro that sits at one point behind GPT-5.6 Sol and a Flash version that is a lot cheaper and faster as well.