The Point of No Return: GLM 5.2 Approaches the Frontier

The Chinese open-weight AI model GLM 5.2 compares with leading models from OpenAI and Anthropic on coding and development tasks. 

Today, we cover the shift toward cheaper AI models, the difference between open weights and closed models, and the U.S. government’s reported ban on Anthropic’s most powerful model after security testing.

------
🌌 LIMITLESS HQ ⬇️

NEWSLETTER:    https://limitlessft.substack.com/
FOLLOW ON X:   https://x.com/LimitlessFT
SPOTIFY:             https://open.spotify.com/show/5oV29YUL8AzzwXkxEXlRMQ
APPLE:                 https://podcasts.apple.com/us/podcast/limitless-podcast/id1813210890
RSS FEED:           https://limitlessft.substack.com/

------
TIMESTAMPS

0:00 China’s Open Source Comeback
3:05 Benchmarks and Cost
8:15 Markets
10:51 A Six-Month Model Gap
13:56 The Fable 5 Ban
15:56 Public Access and Competition
19:26 Open Source vs Open Weights
21:09 Multi-Model Routing Arrives
25:41 Regulation and the Road Ahead
28:31 Closing

------
RESOURCES

Josh: https://x.com/JoshKale

Ejaaz: https://x.com/cryptopunk7213

------
Not financial or tax advice. See our investment disclosures here:
https://www.bankless.com/disclosures⁠

Creators and Guests

Host

Ejaaz Ahamadeen

Host

Josh Kale

What is Limitless: An AI Podcast?

Exploring the frontiers of Technology and AI

Ejaaz:
Last week, a Chinese company released a free AI model that is as good as Anthropik's

Ejaaz:
best model. It also beats ChatGPT 5.5 at writing and coding,

Ejaaz:
but it comes with a twist.

Ejaaz:
It's a sixth of the price and it's completely open source.

Ejaaz:
You can download it and run it at home. Now, in that same week,

Ejaaz:
the United States government banned Anthropik's most powerful model,

Ejaaz:
Fable 5, after someone revealed that an unrestricted version of it had hacked

Ejaaz:
into the National Security Agency's systems.

Ejaaz:
I think we've reached a point of no return. And not to sound dramatic, but

Ejaaz:
in six months, it is very realistic that we will have open source or open weight

Ejaaz:
models that are accessible to anyone in the world with an internet connection

Ejaaz:
and 5 to 10k to run at home,

Ejaaz:
that they can fine tune to do anything.

Ejaaz:
And it's mythos grade level models. These are the same models that we're hearing

Ejaaz:
rumors and reports from verified that they can exploit some of the most secure

Ejaaz:
systems in the world faster than any other exploiter has been able to do in the past.

Ejaaz:
And I think we're going to look back on 2026 as the moment or the year that

Ejaaz:
everything really changed and the point where humanity as itself really needs

Ejaaz:
to focus on safeguards and figuring out how to regulate

Ejaaz:
and release these AI models in the future. So we've reached a convergence of

Ejaaz:
this really interesting trend where the most powerful models in the world are

Ejaaz:
freely available and open source, available for anyone to access.

Ejaaz:
And the government, the United States specifically, has an off switch for their most powerful model.

Josh:
Yeah, it's been a couple of months, it seems, since we've had some news on the

Josh:
frontier of China. And you kind of forget about them every couple of weeks where

Josh:
they just kind of disappear, they quiet down.

Josh:
The new models come out, we see the fables, we see the mythos of the world.

Josh:
But then out of nowhere, they strike back and seemingly every single time it

Josh:
comes as a surprise at how powerful these new models have become so to start

Josh:
with this we have a new model from our favorite company to pronounce jeepu.

Josh:
I feel like i want to name my dog that is such a cute name but jeepu

Josh:
is doing something not so cute they're actually releasing a model named glm

Josh:
5.2 which kind of blew everyone's expectations out of the water i remember way

Josh:
back like six months ago when deep seek was doing this like

Josh:
deep secret release model everyone is like wait you did what with what and

Josh:
that's what this model feels like again we're getting that moment again because

Josh:
this is an open weights model which is not to be confused with open source and

Josh:
we'll talk about that in a little bit but this is an open weights model that is if i'm

Josh:
correct about this within one single point of the sw bench pro benchmark which

Josh:
is the benchmark that a lot of people use for coding oh yeah of gpt 5.5

Josh:
the like frontier coding model from open ai and that comes as a surprise because

Josh:
the cost well one if you run it locally is free but two if you run it on a server

Josh:
is like you said earlier you just one sixth of the cost so you're getting a

Josh:
incredible amount of coding capability for something that costs a fraction of

Josh:
what it costs if you were to go to one of these larger language models and it seems to work,

Josh:
almost as good, if I'm right. And this comes as a surprise to most people because

Josh:
every time we start to count China out, we're like, no, surely they can't catch up.

Josh:
They continue to chip away at this frontier.

Ejaaz:
There's a few things that people will jump to immediately. OK,

Ejaaz:
one, that these benchmarks can be easily gamed.

Ejaaz:
We're going to show you a few examples of benchmarks that couldn't be gamed

Ejaaz:
and GLM 5.2 performs really, really well. But the second thing is the cost.

Ejaaz:
Cost has become a really important point of discussion amongst enterprises specifically that are spending

Ejaaz:
hundreds of millions of dollars per year to access Claude and GPT.

Ejaaz:
It's just too much money for them to spend in terms of like the return on investment

Ejaaz:
that they're getting in work that they actually see.

Ejaaz:
So what they're now turning towards is these free open source models,

Ejaaz:
primarily designed and made by Chinese AI labs that can cut costs down drastically.

Ejaaz:
Just last week, we had Microsoft announce that they're replacing their co-pilot

Ejaaz:
LLM with not ChatGPT, with not Claude, but with DeepSeq itself.

Ejaaz:
So the point is, this comes at a very important time where cheaper models are

Ejaaz:
getting a lot of attention.

Ejaaz:
So now when we look at GLM 5.2 specifically, it is

Ejaaz:
Five to seven times cheaper than GPT 5.5 and Claude Opus 4.8,

Ejaaz:
but performs, as we're seeing on the benchmarks right here, almost as good as

Ejaaz:
each of these models, specifically at the metric that is the most important, which is coding.

Ejaaz:
Now, a lot of skeptics quite rightly were like, I don't know if this is actually

Ejaaz:
true. Like, let me test it against a few other independent benchmarks.

Ejaaz:
It came up pretty high. So if you look at the front end development when it

Ejaaz:
comes to like website design, GLM 5.2 Max is just below Fable 5.

Ejaaz:
We're not even talking about Opus 4.7 or 4.8 anymore, which it absolutely beat.

Ejaaz:
And then when we're looking at like anecdotes or feedback from like distinguished

Ejaaz:
individuals in the Western frontier.

Ejaaz:
So right now we're looking at a tweet from the CEO of Vercel.

Ejaaz:
He goes, I'm genuinely impressed, almost shocked at how good GLM 5.2 is at coding.

Ejaaz:
So this is feedback from real people using this for real use cases.

Ejaaz:
For the last three years, Josh, we've basically been told that the hundreds

Ejaaz:
of billions of dollars that is being spent on AI CapEx is for one single reason

Ejaaz:
only, to gain a moat ahead of any other model provider.

Ejaaz:
So we spend all this money on compute to train a frontier AI model.

Ejaaz:
And that moat, it doesn't matter what other companies do in China,

Ejaaz:
we will have the best model and that's enough for us.

Ejaaz:
This release from Gipu with GLM 5.2 basically shows us the opposite.

Ejaaz:
For a fraction of the cost, you can create a near frontier model that does like,

Ejaaz:
I don't know, 95% of the work,

Ejaaz:
And so it brings into question the valuation between these companies.

Ejaaz:
Should they be spending this amount of money or can we just do it for a lot

Ejaaz:
cheaper like these Chinese AI labs?

Josh:
Yeah, well, the large AI labs, I'm not sure they have a choice.

Josh:
I mean, it's just that you have to continue to push the frontier forward,

Josh:
whether you like it or not.

Josh:
But I think what we're seeing is a lot of these questions that we were excited

Josh:
to see play out, we're starting to get answers to.

Josh:
Like now it's less China versus America and more open source versus closed source

Josh:
because I mean, the open source models are coming from inside too.

Josh:
We have NVIDIA. They're working on open source models that are incredible,

Josh:
and they're making progress in that front.

Josh:
We have Apple now, who has an actually functional Siri on everyone's hardware

Josh:
device that runs essentially for free.

Josh:
So they're slowly starting to nibble away at this, I guess, the lower bottom

Josh:
of the barrel set of use cases.

Josh:
And then we have china which is glm that's deep seek that's these larger models

Josh:
where they're actually competing on the frontier so these big frontier private models are facing

Josh:
heat both from the lower end of the stack but also right at the top where these

Josh:
benchmarks sit and we're going to see how that plays out economically for in

Josh:
the case of jipu at least it's been playing out pretty well and,

Josh:
we probably should talk about the stock a little bit believe it or not this

Josh:
company is publicly traded not here in the united states but this is publicly

Josh:
traded at least in china and it's gone up.

Ejaaz:
What is that

Josh:
1500 percent 15x on the year that's like a crazy return and some interesting

Josh:
facts about this return and it's it's so funny to see kind of i guess how inefficient

Josh:
chinese markets are also note that the chart you're seeing on screen

Josh:
they have a lunch break in their stock market i didn't know this labeled it,

Josh:
like i didn't realize that chinese stock markets had an hour-long lunch break

Josh:
in the middle of the day. So that's cute and that's fun.

Josh:
But the numbers are pretty outrageous. When we trade, when we talk about expensive

Josh:
companies, we talk about SpaceX, who's trading what is it, like a very high

Josh:
multiple towards earnings. And,

Josh:
What we have with Jibu and this company that it's kind of owned by,

Josh:
Knowledge Atlas Technology, it's currently trading at about $136 billion market cap.

Josh:
It made $170 million or $107 million, I should say, in the full year of 2025.

Josh:
That means it trades 1,300 times sales, which is just this unbelievably high

Josh:
multiple on this company.

Josh:
And I think it's a testament to the, I guess, the lack of availability to get

Josh:
AI exposure in Chinese markets, but also the confidence and the excitement and

Josh:
enthusiasm they have around companies like this. That was just an interesting thing to see.

Ejaaz:
Yeah, I mean, at this valuation, it's about, what is that, like a fifth of Anthropics

Ejaaz:
valuation right now, which is, I think, around a trillion dollars.

Ejaaz:
So again, like it begs the question, is Chinese AI labs underpriced or are American

Ejaaz:
companies overpriced? And I'm curious to hear, like what listeners of the show actually think.

Ejaaz:
I tend to think that they probably need to meet somewhere in the middle.

Ejaaz:
We were actually saying before we started recording, Could you imagine the reaction

Ejaaz:
to this news if Anthropic was a publicly traded company and a new 3D open source

Ejaaz:
model that was freely accessible to anyone could achieve pretty much 95%

Ejaaz:
of the capability of Opus 4.8?

Ejaaz:
Like, I wonder what that would have done to the stock price in like a fair market

Ejaaz:
value, but crazy to see nonetheless. So if we're looking at a few different

Ejaaz:
metrics that compare cost and performance, just quickly to run you guys through this.

Ejaaz:
For input versus output tokens, for a million tokens, you're looking at around

Ejaaz:
$1.50 to $4.50 when it comes to cost.

Ejaaz:
Now, comparing that to Opus 4.8, that's around, I believe, $5 versus $25.

Ejaaz:
So again, we're achieving that 3 to 5x cheaper when it compares to a model of

Ejaaz:
similar performance and capability.

Ejaaz:
Now, I was skeptical of the benchmarks, and I have a new favorite benchmark

Ejaaz:
to compare it against, which is called DeepSwee.

Ejaaz:
DeepSwee is basically a benchmark that gives no models any answers.

Ejaaz:
Typically, with a benchmark, you have an answer sheet, and it can kind of cheat

Ejaaz:
and look at it and figure out a way to get to that answer.

Ejaaz:
There's no answer sheet for this

Ejaaz:
one, so it's a very accurate test of how good your model is at coding.

Ejaaz:
For DeepSuite, GLM 5.2 achieved a very modest fifth place. Now,

Ejaaz:
that is probably, or rather, fourth place, fifth place, fifth place.

Ejaaz:
And that is a pretty accurate standing of how agentic coding looks like for

Ejaaz:
this particular model. It is the highest number one place for open source model.

Ejaaz:
It absolutely crushed Kimi K2 by 17 percentage points. or a very clear lead.

Ejaaz:
And it's great to see how it weighs up. Like if it may not be frontier capability,

Ejaaz:
but if you want a workhorse, if you want an agent that basically works overnight

Ejaaz:
and isn't going to break the bank, GLM 5.2 is probably something that you can look at.

Ejaaz:
Another thing is it's really good at front-end web development.

Ejaaz:
So if you're looking at this screen right now, the website that you're seeing

Ejaaz:
was completely one-shotted in about 10 minutes from this one single model, GLM 5.2.

Ejaaz:
And repeatedly across design benchmark, Arena Benchmark was another one that I saw.

Ejaaz:
It performs really highly, in some cases beating Fable 5. So it's a really good

Ejaaz:
front end design model if that is something of interest.

Ejaaz:
And then the final one, because I know a lot of listeners on the show is like,

Ejaaz:
you know, how good are these models at like trading, investing, making money for you?

Ejaaz:
Well, there's this very famous benchmark, which is called the Vending Benchmark,

Ejaaz:
which basically allows an AI model to control a theoretical $10,000 and see

Ejaaz:
if it can make money by stocking a vending machine and then conducting sales,

Ejaaz:
managing inventory against competition.

Ejaaz:
It achieved second place right behind Claude Opus 4.7, which is the current

Ejaaz:
leading model. So it's also pretty good at making money as well.

Josh:
Yeah, and it also has a very clear roadmap to continue to be good and to get

Josh:
even better. There's an interaction actually between Elon Musk and the CEO of

Josh:
Z.ai, who is creating these models.

Josh:
So this guy asked, what's your current timeline for China to reach Fableclass?

Josh:
GLM 5.2 certainly shortened the gap. And then Elon said probably Q1.

Josh:
And then the CEO said, won't take that long. Which means they expect us to get

Josh:
a new Fableclass level model that's open weight and open source within the next six months.

Josh:
Which is incredibly compelling because that is going to be served up as open weights.

Josh:
And as you know, with open weights, you can actually run it on your own hardware.

Josh:
But the question is, do you actually want to run this on your hardware?

Josh:
I see on Twitter all the time, people who are spending tens of thousands of

Josh:
dollars to get those Mac studios, they're stacking them up in their offices,

Josh:
they're trying really hard to run these models locally.

Josh:
And I hate to break it to you, but the math ain't really math in on this so well.

Josh:
So there's a suite by Mike Schweinbach I thought was great. And it says the

Josh:
minimum to run the model is about $20,000 in hardware and you get about 20 tokens per second out.

Ejaaz:
For $20,000, that's like,

Josh:
That's pretty slow. It's not thinking that fast. And if you have these really

Josh:
long chain of thoughts, these long reasoning traces, it's going to take you

Josh:
a very long time to get an answer that involves deep thinking.

Josh:
So for about $20,000, you can get close to 35 billion tokens.

Josh:
And that's a 12 to one input to output ratio, assuming you have like good token caching setup.

Josh:
So he's saying if you ran the hardware 24-7 with zero downtime,

Josh:
it would take roughly five and a half years just to break even.

Josh:
And that right there is why open weights models are incredible.

Josh:
You're probably better off getting it served directly from their servers from

Josh:
the cloud instead of running your own.

Josh:
Because not only do you have to deal with the complexity, you have to power

Josh:
it all on, you have to deal with hardware stuff, and you have to worry about

Josh:
getting the actual hardware.

Josh:
Because Lord knows, getting those computers now is not as easy as it used to

Josh:
be. So interesting note on cost,

Josh:
on how available these are and accessible these are on a relative basis.

Ejaaz:
And the Chinese companies themselves are willing to subsidize these costs, just to be clear.

Ejaaz:
Like to play around with Kimi K 2.7, which is their frontier model,

Ejaaz:
I've been able to access it and use it since they launched it.

Ejaaz:
And I've been free using it to kind of like do research and all that kind of

Ejaaz:
stuff. And I've never once been charged for it. So there's a high subsidy coming

Ejaaz:
from like the Chinese side of things as well.

Ejaaz:
The other thing I'll say is these numbers may look big, right?

Ejaaz:
Like who on earth is spending $20,000 to get hardware that you can like run

Ejaaz:
at home to run these models open source?

Ejaaz:
But the idea is six months from now, 12 months from now, these very same models

Ejaaz:
will be distilled enough.

Ejaaz:
So that means it can maintain its intelligence, but good enough to run on your

Ejaaz:
local hardware at home, a custom PC, or maybe even your laptop.

Ejaaz:
The trend that we're undeniably seeing with these open-source models in particular

Ejaaz:
is higher intelligence for lower-cost hardware.

Ejaaz:
And if that trend continues, we will end up seeing this model that we're talking

Ejaaz:
about today being able to run off your handset. So it's something that seems

Ejaaz:
unfeasible right now to access.

Ejaaz:
But further on down the line, open-source, in my opinion, is pretty undeniable.

Ejaaz:
You'll be able to run it at home, and that's pretty good. But moving on.

Ejaaz:
The reason why we wanted to write this episode is there's a convergence of two trends, right?

Ejaaz:
So last week, we had a lot of reporting around Fable 5 being banned by the United States government.

Ejaaz:
The primary reason is the United States government does not think the model

Ejaaz:
is safe. If placed in a malicious actor's hands, we'll be able to be used against

Ejaaz:
government systems, hack, exploits, all that kind of stuff. And it's proven

Ejaaz:
itself on internal testing.

Ejaaz:
And the most recent revealing was a quote from a senator saying that the head of the NSA

Ejaaz:
Explained in a red team exercise, which is like a controlled environment,

Ejaaz:
that Claude Mythos 5 was able to breach all of its systems.

Ejaaz:
And typically, it would take months for an individual expert to do that.

Ejaaz:
It did it in hours. And this is just a crazy story and headline to read.

Ejaaz:
They've switched it off. It's not accessible to anyone. If you go on cloud right

Ejaaz:
now, you're unable to access Fable 5.

Ejaaz:
But the point is, these two trends have converged at the same time.

Ejaaz:
And it's important to discuss this because very soon in a few months time,

Ejaaz:
as that Elon tweet showed, we're going to end up with Mythos grade level models

Ejaaz:
that are freely available to anyone, subsidized by China or available to run at home for 10k.

Ejaaz:
And that is pretty scary, I guess.

Josh:
Yeah. Is that the lead now? Are we at six months? Does that feel about right?

Josh:
Like if they, if they release Mythos class by the end of this year,

Josh:
and then that gives, I guess, an open AI and Anthropic a six month head start.

Ejaaz:
And then the head of Chippoo has said it.

Josh:
So, yeah. So it seems like that's about right currently where we have like a

Josh:
six month window between us and the current bleeding edge open source.

Josh:
I could see that kind of getting closer and closer. It feels like they're right on the tail.

Josh:
Of course, understanding what's going on internally would be very helpful to

Josh:
know, because I'm sure GPT 5.5, well, we know we're getting 5.6 pretty soon.

Josh:
I'm sure Anthropic is working on something even more powerful than Mythos.

Josh:
And it feels like we don't really have a choice but to continue progressing

Josh:
as fast as we are. Otherwise, these are going to catch up.

Josh:
And they won't have the guardrails that are put in place currently by the Frontier

Josh:
models. Now, what's happening currently is we're seeing this fork.

Josh:
In terms of these private models where only people internally are now able to

Josh:
use them and anyone out in the world is getting, I guess, kind of disabled.

Josh:
They're getting a handicap because they're not actually able to access these frontier models.

Josh:
So we're seeing this weird crossroads where there's a small subset of people

Josh:
that work internally within OpenAI, within Anthropic, that are getting access to these models.

Josh:
The government is limiting their public use, which means the public is getting left behind.

Josh:
And then China is coming up and they're saying, hey, in six months,

Josh:
we're going to be right here at your head.

Josh:
So it's this really interesting dynamic that's at play. And we're going to really

Josh:
have to closely monitor this as these new frontier models continue to be released,

Josh:
because you have to assume, even though the world isn't using Mythos or Fable, they're continuing

Josh:
to iterate and to build better models. They're not just going to stop because of this.

Josh:
Same with OpenAI, same with all the other frontier labs.

Josh:
The question is, are these models

Josh:
going to be held privately for just a small subset of people to use?

Josh:
Or is there going to be this path forward in which the public can use them?

Josh:
I think everyone's hope is that there is a path forward.

Josh:
But currently, we're at this weird standstill where it feels like China's kind

Josh:
of breathing down your neck here.

Ejaaz:
Well, the irony also is if the government is just going to come in and switch

Ejaaz:
off the frontier model, it's going to push companies to use open source models.

Ejaaz:
Imagine you're an enterprise, right? And you're running your entire company

Ejaaz:
on Fable 5 or whatever the frontier model is from an AI lab.

Ejaaz:
And then suddenly you know that the government can just switch the button off

Ejaaz:
and suddenly your company can't do its thing.

Ejaaz:
You're more incentivized to kind of like run an open model at home that's privately

Ejaaz:
inferenced such that you can never shut it down.

Ejaaz:
So if I was an enterprise that has been running Fable 5 and that has now been

Ejaaz:
shut off, I'll be looking over at this GLM 5.2 thing and thinking,

Ejaaz:
well, it's MIT open source.

Ejaaz:
Yeah, maybe it costs 20K to run on hardware, but like I'll rather spend that

Ejaaz:
and save, you know, hundreds of millions down the line versus like going with Fable 5.

Ejaaz:
And yeah, maybe achieving frontier level performance, but then,

Ejaaz:
you know, being shut off potentially by the government, according to their agenda,

Ejaaz:
like that's not something that you potentially want.

Ejaaz:
Now, I want to give a quick counterpoint to the whole Chinese open source AI

Ejaaz:
models are going to take over the world because they're cheaper,

Ejaaz:
they're as good, maybe not as good, but as good, good enough,

Ejaaz:
right? Which is very simple.

Ejaaz:
If you're an American lab that has a frontier AI model that is expensive and

Ejaaz:
you see your neighbors, or if you see your adversaries, China,

Ejaaz:
distilling your model and presenting it as a cheaper model, you just do the same for your own model.

Ejaaz:
And Anthropic has demonstrated that many times, producing Sonnet.

Ejaaz:
Sonnet 4 is basically their cheaper model of Opus 4.8, I believe.

Ejaaz:
And then you see it with ChatGPT, with GPT Flash. These AI labs will produce

Ejaaz:
a cheaper version, and they'll distill it directly from their frontier models.

Ejaaz:
And as these models get good enough to rebuild themselves, it gets easier to do.

Ejaaz:
So I can see a world where they release Fable 6 in the future with a companion

Ejaaz:
model, which is like Sonic 6. And it's super cheap for anyone that wants 85%

Ejaaz:
of the capability and don't care about that extra 15%. And it's super cheap.

Ejaaz:
So it's competitive with the Chinese models. I don't think America has lost

Ejaaz:
the kind of like cheap model argument, but the open source one,

Ejaaz:
they definitely have. I don't see the American and labs open sourcing anytime soon.

Josh:
Yeah, well, we saw MetaPivot very clearly from the open source,

Josh:
but like the savior of the open source world to closed source very quickly.

Josh:
And I mean, that hasn't worked out too well for them or anyone really,

Josh:
which is disappointing.

Josh:
There is a small caveat. Maybe we should cover about what open source actually

Josh:
means because it's not truly open source. There are still some secrets.

Josh:
I think a better way to classify this is open weights. And when you go through

Josh:
training, there's, let's say, a trillion parameters. Each one of those parameters

Josh:
gets tuned over and over and over through each training run,

Josh:
which happens trillions of times.

Josh:
And the output of this are the weights. It's just a large text file that has

Josh:
all of those parameters finely tuned that the model can run off of.

Josh:
What it doesn't include is the actual source code that it took to make that.

Josh:
It doesn't include the ability to reproduce it. All it shares is the outputs.

Josh:
So while you could take their outputs and you could retune and fine-tune those

Josh:
parameters to give you exactly what you want

Josh:
it's not giving you the recipe it's not giving you the secrets on how it built it

Josh:
so there is still some proprietary knowledge as it relates to this open source

Josh:
model these chinese companies because they they are actually preserving the

Josh:
recipe in which they landed on this the data that they trained on there's a

Josh:
lot of secrets the output

Josh:
is what's open source and that's technically open weight so when we say open

Josh:
source i think what we really mean whenever you hear open source model chances

Josh:
are it's open weights and that's a pretty big distinction because that allows

Josh:
them to keep their kind of their secret sauce of how they do it and it's also

Josh:
probably for the better because i assume,

Josh:
you got to imagine they've been distilling some sort of stuff from i mean i

Josh:
remember seed dance that was so like obviously stolen material because it was

Josh:
just able to reproduce all the copyright and video formats from any public tv show in the world so.

Josh:
Where they get their data from leaves a lot to be desired and questioned,

Josh:
but that's kind of the nuance between open source and open weights.

Josh:
And what we're getting right now currently is open weights.

Ejaaz:
I don't necessarily believe it's open models versus centralized models.

Ejaaz:
I think it lands somewhere in between. Now, we've been noticing this new type

Ejaaz:
of product that is getting used by a lot of software engineers and AI users.

Ejaaz:
It's probably best demonstrated by this recent product release from Sakana AI.

Ejaaz:
It's called this new model called Fugu.

Ejaaz:
And they describe it as a multi-agent orchestration system. Basically how it

Ejaaz:
works is you send their model a prompt as you do with ChatGPT or Claude.

Ejaaz:
And it disperses that prompt across many different models. It could be closed

Ejaaz:
models like Claude and GPT.

Ejaaz:
Could be open models like GPO GLM or Kimi K2.7. as well as their own trained

Ejaaz:
model called Fugu, I believe.

Ejaaz:
And the result of this is like agentic debate. So these models kind of produce their own answers.

Ejaaz:
Then you have another model that kind of judges these answers and produces the

Ejaaz:
best answer from all of this.

Ejaaz:
And the result from these tests is basically, not only do you have a better

Ejaaz:
quality output, but it's also cheaper.

Ejaaz:
So the orchestration module basically picks the best models to do something

Ejaaz:
when it's like cheaper, and then only uses the best models when it really needs

Ejaaz:
to solve a really hard task that the other cheaper models can't do.

Ejaaz:
So it saves you a bunch of money, and we see it across other companies like

Ejaaz:
OpenRouter with their new Fusion API. The point being made here is,

Ejaaz:
We are headed towards a world where the ideal AI chatbot uses multiple models,

Ejaaz:
and they may not just be from the same company.

Ejaaz:
So the question I have for the United States government and any government that

Ejaaz:
decides to regulate, whether it's open source models or closed source models,

Ejaaz:
how are you going to regulate every single model in the world,

Ejaaz:
especially when the model labs come from other countries or are in fact open source?

Ejaaz:
You can't regulate open source models. That's the whole idea of it,

Ejaaz:
whether it's open weight or open source.

Ejaaz:
The whole idea is the government can't try to doubt if you're running it on hardware at home.

Ejaaz:
So it's just a really interesting nuance. I just don't think that the stance

Ejaaz:
that the United States government has taken so far is necessarily the most productive

Ejaaz:
one. I understand why they're doing it, but we need to figure out a different framework.

Josh:
It's funny because I saw this news this morning about this Sakana Fugu.

Josh:
I think I'm pronouncing that right. I mean, surely I've never heard of this.

Josh:
I don't know if you've ever heard of this. I think a lot of people watching

Josh:
have never heard of this company. They're Japanese. They came out of nowhere.

Josh:
And suddenly they're posting benchmarks that show that it has higher performance than Fable.

Josh:
And maybe that's true. Maybe they use this mixture of agents.

Josh:
But I think it's also notable that a lot of this is benchmarks.

Josh:
And I actually got some time to play around with the new GLM model this weekend.

Josh:
And while I'm sure it's great at coding and technical use, that's not really

Josh:
what I generally use the models for.

Josh:
And as I'm actually using these models, I'm giving it the general vibe test

Josh:
i'm noticing that i really do strongly bias the american closed source models like,

Josh:
uh gpt and like anthropics um opus and um claude and i mean fable when it was

Josh:
available was incredible and

Josh:
although the benchmarks show that it's very competent at coding a lot of people

Josh:
aren't using it for coding they're using it for other things and

Josh:
and the the general the vibe check doesn't get passed with these models yet

Josh:
at least um so i think that's something worth noting too is like these are just

Josh:
benchmarks i encourage anyone who's listening go try this out for yourself and see for yourself.

Josh:
Some people may actually get a lot of benefit from using a cheaper model.

Josh:
Some people just like having all the context in one place and they want just

Josh:
a better overall experience.

Josh:
With the routing, I think this is a super interesting precedent that we're seeing.

Josh:
Sakana fugu and how they are choosing to route their outputs through a series

Josh:
of open source and closed source models in order to generate a better and more

Josh:
powerful outcome i wonder the costs i noticed that as i was looking through the documentation

Josh:
there was no real cost associated i have to assume it's,

Josh:
not as high but pretty close because it is routing through.

Josh:
A lot of the private models and some open source models in order to get this

Josh:
which means it's probably consuming a good bit of tokens it's not totally going

Josh:
to be this like open source very low price model

Josh:
but it is interesting to see this trend towards more router based applications

Josh:
where not everyone needs to solve this incredibly difficult challenge.

Josh:
Perhaps you spin off a few sub agents, they use a more lightweight model to

Josh:
get you an answer without needing to consume a lot of those higher cost tokens.

Josh:
So it's cool, innovative, I won't say it's novel, we've seen this before,

Josh:
but it's a new iteration of this that is now showing pretty compelling benchmarks.

Ejaaz:
On the cost side of things, if it's anything like OpenRouter's Fusion API,

Ejaaz:
which does the same architecture, it achieves roughly like 30 to 50% cheaper

Ejaaz:
versus the frontier models, which isn't that major compared to like some of

Ejaaz:
the Chinese open source models.

Ejaaz:
But it still saves you a bunch of money if you're an enterprise using this at length.

Ejaaz:
I'm trying to think about the major takeaway that I have for myself after we've

Ejaaz:
done this episode, Josh.

Ejaaz:
And I think the main one is I'm inclined to say, and I hope I'm wrong,

Ejaaz:
that future AI model releases, Fable and above, whether it comes from GPT 5.6

Ejaaz:
or 6 or other frontier AI labs,

Ejaaz:
they're going to be more controlled in their release because governments are

Ejaaz:
going to start getting more involved.

Ejaaz:
We're going to start seeing nationalization attempts from different nation states

Ejaaz:
in order to figure out how to release these AR models because if they're out

Ejaaz:
in the wild, they can exploit and cause some real damage.

Ejaaz:
I don't want to think about what could happen in terms of a major event,

Ejaaz:
but I think we're reaching that point where we need to pay careful attention.

Ejaaz:
So that's what we're trying to do on this episode. At least that's what I'm trying to do.

Josh:
Yeah, I think that's right. Like the speed and acceleration of these models

Josh:
and the cadence in which they're released is up only.

Josh:
If we had a chart that showed you the length of time in between major model

Josh:
releases, It is just getting shorter and shorter and shorter,

Josh:
and that's not changing.

Josh:
So there needs to be a way to reliably be able to push these out.

Josh:
Otherwise, the gap between what exists behind closed doors and what's available

Josh:
to the public is just going to keep growing.

Josh:
And I'm not sure what implications that has, but it sounds like it is noteworthy and something...

Josh:
Something needs to change in a material way because the speed and velocity in

Josh:
which progress is being made is not slowing down.

Josh:
Like, what does this look like a year from now? How quick are these models able

Josh:
to improve themselves? What are the benchmarks look like? Can we even create

Josh:
benchmarks anymore because it will be so capable?

Josh:
We're right on that cusp because we are approaching this vertical asymptote off the curve.

Josh:
And it's just like, it's a little weird. It feels like we're on this roller

Josh:
coaster and we're like kind of going down, but I guess it's inverted where we're

Josh:
going up and we're going up really fast and you're not really sure. It's escaping.

Josh:
It's escaping control in a way well i wouldn't say escaping control but it's

Josh:
just like that it's it's definitely getting fast and it's like okay like if

Josh:
you're driving your car really fast you got to be a little more careful once

Josh:
you reach high speed because like things things can kind of get a little shaky quickly so.

Josh:
We're at that point and models are getting very capable very quickly.

Josh:
I can't imagine what OpenAI's mythos class model looks like.

Josh:
I'm sure they're working on them.

Josh:
We talk about, I mean, the hardware. I always think about these are the Blackwell series models.

Josh:
What happens with the Vera Rubin series models? It's like this,

Josh:
we are going to accelerate so fast. And I think it's important to,

Josh:
yeah, work on these safeguards now where it's still reasonable to catch up,

Josh:
where there's only one model release in which you have to focus on.

Josh:
And there's not 10 different ones from all these different companies that are

Josh:
being pushed every single week so interesting that's the update china is,

Josh:
back with their open weights model not to be confused with open source and um

Josh:
yeah we still don't have fable access so

Josh:
hopefully these things will get sorted but i think it's it's noteworthy that

Josh:
china they they never disappeared i want to know what deep seek is doing next

Josh:
i think that's my next question is like where's deep seek at where's deep seek v v5 or v6 they just.

Ejaaz:
Raised a massive round 50 billion dollars um that's their valuation at least

Ejaaz:
there's still a fraction of frontier labs but yeah they raised like uh was it

Ejaaz:
nine billion dollars the founder himself put in three billion dollars there

Ejaaz:
they're doing pretty well and we haven't seen a model race from them anytime soon

Josh:
Yeah yeah so will be fun to see but that is the update on china on open source

Josh:
thank you guys so much for watching as always if you enjoyed this episode don't

Josh:
forget to share it with a friend who might also like the show who might care

Josh:
about china or open source models or wherever it may be

Josh:
if you listen on a podcast player rating us how you believe we deserve to be

Josh:
rated is always appreciated we

Josh:
love the five stars those are always great uh newsletter twice a week next one

Josh:
is dropping on wednesday a day after you listen to this and yeah that's i have one final.

Ejaaz:
Request josh something that you and i discussed on our on our walk uh last week but um

Ejaaz:
We are in the market for sponsors or anyone that can support us, please.

Ejaaz:
Josh and I and producer Luke have been keeping the lights on this entire time

Ejaaz:
and we've reached a point where we're feeling really confident about the numbers

Ejaaz:
and all the support that you guys have given us.

Ejaaz:
And we would love to have a partner that we feel very passionate about join

Ejaaz:
us and support us in our vision of growing this into the leading frontier and

Ejaaz:
AI tech podcast in the world.

Ejaaz:
So if there's anyone out there listening to this that is inspired or wants to

Ejaaz:
support us, let us know, DM us, you know, we're on X, we're everywhere,

Ejaaz:
just reach out and we would love to hear from you.

Josh:
That would be great. All the support is very much appreciated.

Josh:
Keep the lights on around here and keep things going strong.

Josh:
So yeah, thank you as always for the support. If you made it this long,

Josh:
you're a real one and hopefully you enjoyed this episode. So thank you as always

Josh:
and we will see you on the next one.