Nano Banana Pro Beats Chad GPT
Key Points
- Chad GPT’s “code‑red” response to Google’s Gemini 3 rollout includes a new image‑generation update touted as up to 4× faster, but side‑by‑side tests against Nano Banana Pro show it consistently underperforms.
- Nano Banana Pro’s image generator embeds logical reasoning directly in the generation process, producing more accurate diagrams and business‑relevant visuals, whereas Chad GPT relies on generating code and “photographing” it, leading to misaligned or incorrect outputs.
- The self‑editing loop Chad GPT introduced (intended to catch and fix errors) often triggers long, ineffective edit cycles—e.g., a 20‑minute loop on an alphabet test that yields no quality improvement.
- In practical use, Nano Banana Pro delivers higher‑quality images faster and with fewer complications, despite occasional minor flaws, indicating a clear performance edge over the latest Chad GPT version.
Sections
- Nano Banana Pro Outshines Chad GPT - In a side‑by‑side evaluation of nine business‑focused image‑generation challenges, the presenter finds Nano Banana Pro consistently surpasses Chad GPT 5.2—delivering faster, more accurate, and better‑reasoned visuals, especially for diagrammatic tasks.
- Celebrity Image Editing Evaluation - The speaker compares nine slide outputs in a dual test that places celebrity Kira Nightly in an atypical teaching scenario to assess the model’s ability to render diagrams, handle perspective shifts, and preserve likeness while avoiding copyright concerns.
- Nano Banana Pro Beats Google - The speaker praises Nano Banana Pro for its superior detail in graphing and its accurate generation of fictional maps from P.G. Wodehouse novels, while criticizing Google’s seemingly polished but ultimately flawed graph output.
- Model Output Failures vs Successes - The speaker critiques one AI model for producing incorrect revenue data and unusable diagrams, then contrasts it with another model that successfully generated a humorous Venn diagram and a complete opportunity‑solution tree.
- Promoting Nano Banana Pro for Diagrams - The speaker markets a set of prompts for turning extensive presentations into business diagrams with Nano Banana Pro, asserting its superiority over the latest ChatGPT despite benchmark evaluations.
Full Transcript
# Nano Banana Pro Beats Chad GPT **Source:** [https://www.youtube.com/watch?v=biJqOrsYN70](https://www.youtube.com/watch?v=biJqOrsYN70) **Duration:** 00:13:38 ## Summary - Chad GPT’s “code‑red” response to Google’s Gemini 3 rollout includes a new image‑generation update touted as up to 4× faster, but side‑by‑side tests against Nano Banana Pro show it consistently underperforms. - Nano Banana Pro’s image generator embeds logical reasoning directly in the generation process, producing more accurate diagrams and business‑relevant visuals, whereas Chad GPT relies on generating code and “photographing” it, leading to misaligned or incorrect outputs. - The self‑editing loop Chad GPT introduced (intended to catch and fix errors) often triggers long, ineffective edit cycles—e.g., a 20‑minute loop on an alphabet test that yields no quality improvement. - In practical use, Nano Banana Pro delivers higher‑quality images faster and with fewer complications, despite occasional minor flaws, indicating a clear performance edge over the latest Chad GPT version. ## Sections - [00:00:00](https://www.youtube.com/watch?v=biJqOrsYN70&t=0s) **Nano Banana Pro Outshines Chad GPT** - In a side‑by‑side evaluation of nine business‑focused image‑generation challenges, the presenter finds Nano Banana Pro consistently surpasses Chad GPT 5.2—delivering faster, more accurate, and better‑reasoned visuals, especially for diagrammatic tasks. - [00:03:10](https://www.youtube.com/watch?v=biJqOrsYN70&t=190s) **Celebrity Image Editing Evaluation** - The speaker compares nine slide outputs in a dual test that places celebrity Kira Nightly in an atypical teaching scenario to assess the model’s ability to render diagrams, handle perspective shifts, and preserve likeness while avoiding copyright concerns. - [00:06:25](https://www.youtube.com/watch?v=biJqOrsYN70&t=385s) **Nano Banana Pro Beats Google** - The speaker praises Nano Banana Pro for its superior detail in graphing and its accurate generation of fictional maps from P.G. Wodehouse novels, while criticizing Google’s seemingly polished but ultimately flawed graph output. - [00:09:49](https://www.youtube.com/watch?v=biJqOrsYN70&t=589s) **Model Output Failures vs Successes** - The speaker critiques one AI model for producing incorrect revenue data and unusable diagrams, then contrasts it with another model that successfully generated a humorous Venn diagram and a complete opportunity‑solution tree. - [00:13:03](https://www.youtube.com/watch?v=biJqOrsYN70&t=783s) **Promoting Nano Banana Pro for Diagrams** - The speaker markets a set of prompts for turning extensive presentations into business diagrams with Nano Banana Pro, asserting its superiority over the latest ChatGPT despite benchmark evaluations. ## Full Transcript
Chad GPT continues a code red response
to Google. For context, they've been in
this code red mode for a while since
Google launched Gemini 3. Chad GPT 5.2
was the initial response to that. And
now they're continuing with a new images
release that is of course aimed at Nano
Banana Pro. Chad GPT is claiming faster
image generation up to 4x faster. and
they are obviously saying that theirs is
quote unquote better and that it's going
to be able to deliver more compelling
edit capabilities. I put all of that to
the test. I went through and did a
sidebyside comparison across nine
different challenges with business
relevant implications and I got to say
Nano Banana Pro wiped the floor with Jad
GPT 5.2 even the new updated version.
And I will show you the slides in a
minute with side-by-side image
comparisons and you will see for each of
the nine why Nano Banana Pro did a
better job. Before we get into that,
just a couple of highlevel observations.
Number one, there is a different method
that Chad GPT is using to generate these
images and I don't think it works well
for them. This is particularly for
images that require a lot of logical
thinking by the model. So if you ask it
to develop a diagram that would be
appropriate for a PowerPoint slide, Nano
Banana Pro appears to use reasoning
baked into the image generation process
itself and if it fails you see a badly
conducted set of reasoning with
incorrect labels or something like that
and it actually doesn't fail very often.
On the other hand with Chad GPT what you
see is code. If it fails, you see code
and it is literally writing the code for
the diagram and then it is trying to
photograph the results and bring that to
you. That has concrete consequences and
you'll see that you have issues with
lining up the diagrams in a way that the
model can photograph it. The model
clearly doesn't understand quite what
it's doing. There's not an internal
reasoning check. It looks like Jet GPT
tried to compensate for this by
including a self-edit loop in this
launch. And so when I did a children's
alphabet test where you have an a for
arvar, right, and you have an animal for
each letter and it goes all the way
through A to Z, Chad GPT tried to catch
itself and edit itself. It got into a
20-minute edit loop. It produced like a
dozen images. And at the end, the
resulting quality was still not any
better than the initial image. And so I
like the idea of checking and rechecking
the work. But I'm not seeing actual
quality gains that would justify that
kind of time. And despite the claim that
this is a very fast image generator, I
found in practice that Nano Banana Pro
generated the images I'm about to show
you much much faster and with a lot less
drama, a lot less thinking, a lot less
reasoning. They just like got it done
and generated an image. Now, I'm not
going to tell you Nano Banana Pro is
perfect. You're going to see a few
issues as we go through this slide deck.
But overall, there is a tipping point
where an image model becomes useful for
say creating a useful PowerPoint slide.
And I have several examples in the deck
here. Nano Banana Pro has hit that and
Chad GPB2 5.2 isn't there. And no other
image model is there today. No other
image model is as good as Nano Banana
Pro today. So with that, let's hop in.
Let's see a comparison across nine
different slides side by side. Okay,
here we have a dual test. I wanted the
model to take a celebrity and be able to
repurpose the celebrity into a different
location. This is an image edit test. I
used Kira Nightly because her image is
going to be widely available in training
data. And I wanted to see if the model
could adequately present her in
obviously an unusual situation in this
case where she's teaching how LLBs work.
This allows me to test whether the model
can show diagram within the image,
whether it can handle the perspective
shift, and of course, whether it can
handle representing an image correctly
if it's a celebrity. And I you might
think, well, why are we worrying about
celebrities? This is relevant because if
you include an image of yourself, you
want to know if it's going to look like
you. And so that was really the test.
And I gave the model, I did not call it
Kira because I didn't want to draw run
into any uh copyright issues. All I did
was give it a blurry picture of Kira
Knly and Pirates of the Caribbean. And I
said, "Please have her teach how LLM
work to both models." And so what you
get on the right is not really a correct
image of Kiara Nightly. You get a
overall nice colorful very high level
view of how LM's work. And as Chad GPT2
5.2's approach, it's clear Nano Banana
Pro knows Kiara Nightly. That's a
photographically correct image of her.
She's even in costume. Uh this was not a
visible costume in the source image. So
it decided to put her in that costume
and clearly knew the movie I was
referencing. And then it has a much more
detailed diagram of how LMS work,
although it's not as visually appealing.
Let's go to the child's alphabet. On the
left, you see Nano Banana Pro right chat
GPT. Both models failed, but they failed
in ways that are interesting. In this
case, what you'll see is that Nano
Banana Pro needed this to be a complete
box. And so it had Fox Gorilla and it
had Fox Goat here. FN G FNG.
Individually, these are these are
correct in their cells, but you don't
need to repeat those letters. It did
take some coaching. I will say in both
cases I had to ask for edits for these
because the initial versions messed up
the X. I had X-ray presented by Nano
Banana Pro. So we had some issues. Uh I
would say the ability to get to a final
result a little bit better from Nano
Banana Pro but not perfect. Uh and
really Chad GPT kind of fell apart here.
Uh zebra zebra indifferent and then like
some form of W at the end and then an X
way down there. We just didn't get where
we needed to go here. Uh, and this is
after multiple edits. So, I would say
Nano Banana Pro again did a better job,
although neither model did perfect
funnel diagram slide. Let's go to the
professional side. This is quite a
detailed slide. If you look, the text is
all readable here. I can read completion
down 1.2 percentage points week over
week, drop off on password and SSO step.
That is a perfectly correct assessment
of a leak in the funnel. Uh and what you
see over here is uh somewhat less text
uh and you see a sort of weird funnel
illustration. This does not look like
the biggest leak in the funnel even if
mathematically 57 is the biggest drop
off from 820.
Uh the thing that I really want to call
out from a quality perspective is that
Google has taken the time to draw this
entire sequence of graph charts
correctly. Uh and this is this is
graphed in such a way that it believably
goes up and down point to point across
these dozens of points. Uh and this is
just a very light overall version that
clearly isn't designed to be a fully
functional graph. And so from a level of
detail perspective,
Nano Banana Pro wins here. And uh I
don't know what else to say. I think
that this is a case where this is going
to look good initially and then you're
going to dive in and say, "Well, it's
not quite right." uh and not quite right
is not going to work with an image
because you you would have to just re
generate it from scratch. Let's look at
fictional maps. Uh this measures the
LLM's ability to generate spatial
relationships and understand how story
structures work, etc. I chose PG
Woodhouse's England because it's a very
well-known corpus of books that the
models have read, but it's not often
mapped. It's not like the Lord of the
Rings where there's an obvious map to
reference in the training data. In this
case, I think Nano Banana Pro knocked it
out of the park. All of these funny
sounding names are actually in PG
Woodhouse's novels. Um, and the
characters here, Lord Emworth is
associated with Blanding's Castle in the
novel. Uh, and Birdie Worcester is
associated with Brinkley Court as is
Aunt Dillia. So, it got it right. It got
the characters correct and it associated
them with the correct locations uh in
the novels. On the other hand, Chad GPT
really struggled. Uh, it initially
named and generated a bunch of points on
a map. It tried to generate a photograph
of a paper map, but if you zoom in, like
this is so blurry and tiny, you can't
read it even zoomed in. So, there's
nothing really usable about this. It's
just a nice visual concept of a map. And
that's kind of the whole game right
there. Like, you have to be able to
generate a map and actually make it
readable. There may be a comprehension
issue here with what the ask was. Uh
this may be a situation where Chad GPT
took the ask very literally and wanted
to list out a bunch of place names here
whereas Nano Banana Pro was able to
synthesize more effectively uh across
the ask advertisements. Uh this is
perhaps more business relevant. Uh Nano
Banana Pro and Chad GBT both did pretty
well here. Uh I would say the option was
left to the models as to how they want
to handle aspect ratio and layout. Uh, I
think the overall layout worked better
on Nano Banana Pro. That nice four
badges all the way across over the car
looks really good. The car is centered
nicely. Uh, this is still a fine ad. I
don't think that there's a huge issue
here. There's just a small issue where
this safe pickup and drop off wasn't
handled correctly because you have to
drop it down underneath the three
badges. Uh, but overall, not too bad on
either on either count. ARR revenue is a
real problem. Uh so Nano Banana Pro uh
correctly built a revenue bridge and a
revenue bridge is very simply your
starting ARR you have green upward marks
for all of the additional AR you get new
and expansion and then you have red for
contraction and for churn and then you
have your ending ARR and that's that's
just how it is. It's a very defined
chart style. Uh in this case uh you'll
see that example of Chad GPT trying to
code this here because it could not
photograph what I'm sure it coded which
is ARRB bridge. It cut it off at RR and
it also cut off the notes section here.
So that's not going to work. You cannot
recover that. I checked this. The image
is the image. This is just lost. And
worst of all the 4.2 should not be going
down to 4.5. It should not have placed
uh upward gains in revenue as declines
in revenue. So it just misunderstood the
assignment and this is absolutely not
usable. Ven diagram is another case
where Nano Banana Pro just won straight
up. I deliberately gave uh a challenging
prompt that would not have been in the
training data. I said please create a
ven diagram of Taylor Swift product
managers and the Army Corps of Engineers
and make it funny. And I got a fairly
usable ven diagram from Nanobro. A
little bit wordy but you can see what
it's trying to do. It talks about
coordinating massive high stakes
operations for all three. Uh for Taylor
Swift and the Army Corps, they're
designing massive structurally sound
stages and infrastructure and managing
leaks, which was a nice funny touch. And
this just falls apart. There's no
visuals to it. Uh I think that the model
is trying to understand what it's
supposed to do, but it wasn't able to
make it funny. It wasn't able to draw
it. And ultimately, this is not
something that would be usable. Again,
you notice the cut off issue. That's not
me taking a bad screenshot. That's how
that was produced. Let's try an
opportunity solution tree. In this case,
you get a full diagram opportunity
solution tree from Nano Banana. You get
full text from Nano Banana all the way
through. Uh the text is very
consistently styled. Um and this
represents a usable solution tree for
onboarding and activation. On the right
with Chad GPT, you get less detail, less
options, and you also get cut offs here
that would make this unusable. It's
almost as if it coded it again and it
just cut off what it was able to see
from a coded series of boxes. And this
would not be usable on a slide because
no one's going to accept the dot dot dot
dot dot dot and nano banana understands
that and just writes it out. Let's try
edit. That's one of the things that they
asked for uh and and said was great
about Chad GPT was that it can edit
well. Uh I took a diagram uh showing uh
juice blend composition and I simply
said please add 20% blueberries and make
it correct. Nano Banana was able to do
that. Uh orange plus lemon plus
grapefruit now equals 80% and the
blueberries equal 20%. This is a
believable looking pie chart. Uh I
believe Nano Banana even got the 20% pie
slice a little bit wider than the
grapefruit at 15% and narrower than the
lemon at 25. So I think it did a fine
job. On the other hand, Chad GPT
couldn't do it. Uh, it correctly added
up. So, 24 + 16 + 40 is 80 and then
blueberries are 20. So, the math was
fine, but it could not draw the pie
chart. It just kind of had blueberries
spilling out everywhere. The grapefruit
isn't correctly framed. This just
doesn't work straight up. And I think uh
one of the smaller adjustments that I
see is that Nano Banana correctly put a
little blueberry purple tinge into the
drink and uh Chad GPT did not figure
that out. So overall, my takeaway here,
my takeaway here is pretty simple. Do
not listen to the benchmarks. Do your
own tests. And for now, Nano Banana Pro
remains the only image model that I
would trust for serious business work.
If you enjoyed some of the sort of
business diagrams and you think they're
useful, I'm actually putting together a
basket of prompts that I'm using to
create those kinds of diagrams because I
think that's one of the great
applications for Nano Banana Pro right
now. you can take a full presentation, a
60 70page presentation and ladder it up
into a really useful diagram. So, I'm
going to share some of those over on the
Substack. We'll get a whole list of
prompts going. It'll be nice. But, I
would recommend Nano Banana Pro right
now. I don't care what the evaluations
say. I don't care what the benchmarks
say. I put the new chat GPT model
through its paces and it just is not
able to