Truly looks like we are getting
extremely close to the launch of
OpenAI's new flagship model, GPT 5.6.
And we're starting to see massive AB
testing across the arena and even inside
ChatGPT with two different checkpoints
appearing in testing. And we now have
what it looks like a confirmed planned
launch date, which is next Thursday,
June 25th. Which is insanely close.
What's interesting is that OpenAI may
have chosen a weaker version of the GPT
5.6 checkpoint, which is the Kindle
Alpha instead of Kepler Alpha, which
many people thought was the stronger
version. But if you're wondering how
people on the timeline seem to be
accessing the GPT 5.6 Pro, it looks like
it is being stealth tested and GPT 5.5
Pro is selected in ChatGPT at least for
some Pro accounts.
So, if you have a Pro account, what you
can do is simply go ahead into ChatGPT,
you can select GPT 5.5, and you want to
set it to Pro. And whenever you have set
it to Pro and type in anything like
create me a landing page, it is going to
give you either two options. You'll
either get a checkpoint version of the
generation or you'll get the GPT 5.5
Pro. And essentially, this is a way for
you to test out the new GPT 5.6 that is
being tested right now. But what we know
so far about the GPT 5.6 Pro is already
pretty interesting cuz the new model
reportedly has a higher juice value for
reasoning effort budget of 960 compared
to the previous model of GPT 5.5, which
was listed at 768.
This is suggesting that GPT 5.6 Pro may
be able to think longer, plan deeper,
and handle more complex agentic tasks.
Now, we already know the GPT 5.5 on high
thinking mode is the best model that is
available, obviously before Fable 5 was
released. But we're also hearing that
with this new release, it would take GPT
to a next level in terms of the thinking
process that would be undertaken
whenever it's prompted to think deeper
on different tasks, and that way you can
get various sorts of tasks done at an
extremely high level. What What we're
also hearing about the model is that it
has a knowledge cut-off that has been
moved to December 2025 compared to the
previous August 2025 cut-off on earlier
versions. On top of that, a new image
was provided showing a different
reasoning style, which could point to a
change to how the model structures its
thinking and responses. But, the biggest
part though is tool integration. They
write appears to be integrated within
ChatGPT and within this model. And
browser use is also something that has
been built in, which could make GPT-5.6
Pro much stronger for real-world agent
workflows, web automation, testing,
research, and coding tasks. Building a
voice agent used to mean gluing together
four different services. One for
speech-to-text, one for the large
language model, one for text-to-speech,
another for turn detection. That means
four plus dashboards, four invoices,
latency issues, and the classic problem
where the agent mishears a phone number
or cuts you off mid-sentence. Well,
AssemblyAI fixes this with their new
voice agent API. It collapses the full
stack into one web socket, which means
speech recognition, large language model
reasoning, voice generation, turn
detection, tool calling, session
resumption, all in one. It's powered by
Universal 3 Pro streaming, so it's built
to handle the hard stuff like names,
>> [music]
>> accents, account numbers, and
confirmation codes.
The turn detection is also speech-aware,
so the agent understands when you're
pausing versus when you're actually done
talking. And the The is simple. $4.50
per an hour flat. That means a 5-minute
call costs around 38 cents only. So,
instead of wiring together a bunch of
services, you can build and ship a
working voice agent today. I actually
built a quick demo for AI builders and
you can try it using the link in the
description below for free with no
credit card. Thanks to AssemblyAI for
sponsoring today's video. Now, I will
say front-end wise, the model has been
drastically better from its previous
version and you can see that with a
couple of these different generations,
but I still believe it's still not there
at the same quality as what we're
expecting with the GPT 5.6. Now, this
SpaceX clone that it was able to
generate was thoroughly and executed
quite well with this generation, but
this is something that is only done
based off of giving it a more detailed
prompt and you can see that it does keep
a strong visual hierarchy. The
production quality is exceptional for
the front-end code and you can see that
it was able to output polished UIs with
scroll triggers and all the components
that you would expect for a landing
page. So, overall, it does do the job
perfectly, but when you are prompting it
in general cases, you will notice that
the packages are still outdated and it's
not something that we would expect from
the GPT 5.6 if you understand what I'm
trying to say. But then again, guys,
when you are to compare the new GPT 5.6
checkpoint to the current GPT 5.5, you
will notice that there is a substantial
gain and they did try their best to get
rid of that GPT style UI components that
is usually generated with each front-end
and that is something that is being
displayed right now with this comparison
and I truly believe that it has gotten a
lot better, but there is still some work
left to do for these new GPT models.
And don't get me wrong, I truly believe
that this model is exceptional with
front end. It's just the taste is not at
the same quality as Opus, and I would
still put it behind it by
a pretty big margin, I would say, cuz it
does
do better with token consumption, but
still regardless, it takes a lot longer
to get that output, and you're not
getting the same sort of quality as the
Opus models in terms of front end
design.
After Fable 5's generation of the
Minecraft clone, this has to be ranked
number two.
Following the Fable 5, and this is just
exceptional, guys, cuz you have a
full-on village that was generated. You
even have villagers, different mobs. You
even have the break animation, which is
something that most models don't
actually do. You have torches, which
emit light, so that is actually pretty
unique on its own. You even have a
crafting system, and what's also nice is
I believe they have added in a cave
system directly into this Minecraft
clone. So, we are to dig a bit deeper,
you will notice that there's actually a
Minecraft cave system that has been
generated as well. So, this is just
exceptional. The fact that it was able
to do this is quite impressive. You have
ores, you have different sorts of
blocks. You have the ability to easily
mine it, but there's no actual
functionality as in like what we saw
with Fable, which is where you have a
crafting system that develops all of the
tools that you would basically see
within Minecraft, but that hasn't been
actually generated with this clone.
Regardless, we do have a pretty good
decent clone that was generated. As you
see, I'm falling into the lava, and I am
taking damage. So, overall, the main
functionality of the game has been
generated, so it did do a pretty decent
job in creating the sandbox game, and I
would rank it number two following Fable
5. There's a lot of stuff in the AI
space that I don't really put on to the
YouTube channel, and you can actually
access it through my free newsletter
with the link in the description below,
where you can subscribe completely for
free. Something that I have also noticed
is that the GPT-5.6 Pro looks excellent
for voxel, meaning that it is able to
generate different sorts of 3D space
generations quite well with different
volume profiles, proportions,
composition, material, lighting, and
animation all in a coherent manner. And
you can see that with this voxel art of
the Om Nom, which is from the Cut the
Rope game. And you can see that it did a
pretty great job with the ambience. You
have even a blinking animation for this
creature. So, that is actually pretty
nice. On topic of voxel, the GPT-5.0 Pro
is truly impressive cuz in just 30
minutes it generated this ultra-detailed
voxel rocket that is complete with a
launch mechanism. You have dynamic
follow cameras, visual effects, and even
procedural generated sound effects,
which you can hear right now in the
background. But, what's the most
impressive is that the level of
cohesiveness between the systems. The
visuals, the physics, the camera work,
the audio, and the overall presentation
all feels like they were designed work
together rather than just being stitched
together piece by piece in a random code
generation. So, overall I got to say
this is just truly remarkable that the
GPT-5.6 is able to output something like
this. And it is in many cases even
outperforming the Fable 5 with these
generations. Here, GPT-5.6 Pro was able
to generate an entire game in one shot,
in a single HTML file. And this is where
it recreated the Sim game, where you
have the ability to build different
sorts of houses. And you can even
control over multiple simulations.
Relationships and social interactions
have been thoroughly generated as well.
You have an autonomous AI with needs and
emotions, like how it is within the
actual game. They have different
careers. They make money. All of these
different things are progressing
simultaneously. Obviously, it's not
picture perfect in a lot of different
ways, but still the fact that it's able
to incorporate all the functionalities
from the real game is nice cuz you can
see that you're able to build different
houses. You're also able to create
different careers, and you even have
random events and weather changes
throughout this game. So, this is just
insane cuz we are seeing that the GPT
5.6 Pro is looking seriously strong for
complex simulations and how it's able to
deploy different agents to have various
sorts of behaviors and how it's able to
use those components to build out
anything you requested. And when you are
to compare the GPT 5.6 to GPT 5.5, you
will definitely notice that there is a
substantial gain in quality. Now, I am
really curious to see how the pricing is
for the GPT 5.6. Hopefully, it is not a
substantial gain from the predecessor,
and I really hope that it is something
that is also as efficient as the GPT 5,
which I do believe will be, but just
kind of scared. Now, take a look at this
Minos 11 operating system. Can you tell
that this was actually something that
was generated in SVG?
It's kind of hard cuz this is something
that was outputted by the GPT 5.6 Pro,
where it was requested to create a
Windows 11 operating system in SVG. And
honestly, I definitely see this better
than Fable 5 in SVG cuz you can see that
it did quite accurately well in almost
all of the components. Obviously, there
are certain things that it didn't do,
and this is something that does better
than Fable 5 in a lot of these cases,
but we're getting to a point where GPT
is getting exceptionally well at SVG,
where it may even be the best model to
use for SVG code generation. 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. If you like this video and would
love to support the channel, you can
consider donating to my channel through
the Super Thanks option below. But,
that's about it, guys, for the GPT-5.6
leaks. Hopefully, I'm back before it
actually drops, which most likely I will
be, but still, I will keep you guys
posted on X, Discord, as well as on
YouTube. I mean, I can sure I can make a
lot of these leak videos to keep you
guys up-to-date on what's actually
happening with this model. But,
regardless, hope you enjoyed today's
video and got some sort of value. This
is going to be another remarkable model
from OpenAI, and it is looking like
OpenAI is finally back. It may not be at
the same level as Fibble 5 yet, but
still, we're starting to see some huge
progress from these GPT models. So,
definitely nice to see from OpenAI. But,
with that out, guys, hope you enjoyed
today's video and got some sort of
value. I'll leave all these links that I
used in today's video in the description
below. Make sure you go ahead and take a
look at the World of AI 5 coding
benchmark tool that I have made, and
I'll leave a link to that in the
description below. Make sure you also go
ahead and take a look at the second
channel, join the newsletter, join the
Discord, follow me on Twitter, and
lastly, make sure you guys subscribe,
turn on the notification bell, like this
video, and please take a look at our
previous videos so that you can stay
up-to-date with the latest AI news. But,
with that out, guys, have an amazing
day, positivity, and I'll see you guys
fairly shortly. Peace out, folks.