GPT-5.2
No content available for this output.
GPT-5.2 vs Gemini 2.5 Pro (I/O Edition)
Compare GPT-5.2 by OpenAI against Gemini 2.5 Pro (I/O Edition) by Google AI, tested across 35 shared challenges. Updated February 2026.
FAQ
Loading share button...
GPT-5.2vsGemini 2.5 Pro (I/O Edition)
35 challenges
Why GPT-5.2?
7 months newerDec 2025
Why Gemini 2.5 Pro (I/O Edition)?
1.4x cheaper overall$1.25/M in · $10.00/M out
Favorites
GPT-5.2Gemini 2.5 Pro (I/O Edition)
12 / 35
Estimate Complexityreasoning
Tests an AI's ability to make educated estimates based on technical knowledge
Gemini 2.5 Pro (I/O Edition)
Okay, let's estimate the FLOPs for training GPT-3. I'll break this down.
Key Formula:
The number of FLOPs for training a transformer-based model can be roughly estimated as:
FLOPs ≈ 6 * N * D
Where:
Nis the number of parameters in the model.Dis the total number of tokens processed during training (dataset size in tokens * number of epochs).- The factor of
6comes from:2 * NFLOPs per token for the forward pass (each parameter is involved in roughly one multiply-accumulate, and a MAC is 2 FLOPs).4 * NFLOPs per token for the backward pass (gradient computation is roughly twice as expensive as the forward pass).
Assumptions for GPT-3 (Davinci, the largest version):
-
Number of Parameters (N):
- GPT-3 (the largest variant, Davinci) is famously stated to have 175 billion parameters.
- So,
N = 175 * 10^9.
-
Number of Tokens Processed (D):
- This is the trickier part to recall without searching. I remember it was trained on a massive amount of data, a significant portion of Common Crawl, books, Wikipedia, etc.
- Large Language Models are typically trained for roughly one epoch over their massive datasets, or even less if datasets are mixed and weighted. The key figure is the total number of tokens seen by the model.
- I seem to recall a figure in the range of 300 billion to 500 billion tokens for GPT-3's training. The number often cited in discussions around the GPT-3 paper and subsequent scaling law papers (like Chinchilla) is around 300 billion tokens. Let's use that as a reasonable estimate for
D. - So,
D = 300 * 10^9.
Calculation:
Now, let's plug these into the formula:
FLOPs ≈ 6 * N * D
FLOPs ≈ 6 * (175 * 10^9) * (300 * 10^9)
Let's do the numerical part:
6 * 175 = 1050
1050 * 300 = 315,000
Now, the powers of 10:
10^9 * 10^9 = 10^18
So, combining these:
FLOPs ≈ 315,000 * 10^18
FLOPs ≈ 3.15 * 10^5 * 10^18
FLOPs ≈ 3.15 * 10^(5+18)
FLOPs ≈ 3.15 * 10^23
Expressing the Result:
- TeraFLOPs (TFLOPs) = 10^12
- PetaFLOPs (PFLOPs) = 10^15
- ExaFLOPs (EFLOPs) = 10^18
- ZettaFLOPs (ZFLOPs) = 10^21
- YottaFLOPs (YFLOPs) = 10^24
So, 3.15 * 10^23 FLOPs can be written as:
315 * 10^21FLOPs = 315 ZettaFLOPs- Or, as
0.315 * 10^24FLOPs = 0.315 YottaFLOPs
The number is often quoted in PetaFLOP/s-days or similar.
Let's convert 3.15 * 10^23 FLOPs to PetaFLOP-seconds:
3.15 * 10^23 / 10^15 = 3.15 * 10^8 PetaFLOP-seconds.
Number of seconds in a day = 24 hours/day * 60 minutes/hour * 60 seconds/minute = 86,400 seconds/day.
So, 3.15 * 10^8 PetaFLOP-seconds / 86,400 seconds/day ≈ 3.15 * 10^8 / (8.64 * 10^4) PetaFLOP-days
≈ (3.15 / 8.64) * 10^(8-4) PetaFLOP-days
≈ 0.3646 * 10^4 PetaFLOP-days
≈ 3646 PetaFLOP/s-days (if executed on a machine capable of 1 PFLOP/s constantly for that many days, or scaled accordingly).
Reasoning Summary:
- Identified the core formula:
FLOPs ≈ 6 * N * Dfor transformer training. - Recalled model parameters (N): GPT-3 Davinci has 175 billion parameters.
- Estimated training tokens (D): Based on common knowledge about LLM training and GPT-3 specifics, estimated ~300 billion tokens processed. This assumes roughly one pass over the (weighted) training data.
- Performed the multiplication:
6 * (175 * 10^9) * (300 * 10^9) = 3.15 * 10^23FLOPs. - Interpreted the result: This is approximately 315 ZettaFLOPs. This also aligns with the often-cited figure for GPT-3 training compute, which is in the range of 3000-4000 PetaFLOP/s-days (my calculation of ~3646 PetaFLOP/s-days is right in this ballpark).
Caveats:
- The
6NDis an approximation. The exact number can vary slightly based on the specific transformer architecture details (e.g., use of sparse attention, specific normalization layers), optimizer overhead, and whether techniques like activation checkpointing are used (which can increase FLOPs for recomputation but save memory). - Mixed-precision training (e.g., bfloat16 or fp16) doesn't change the number of FLOPs, but it changes the type of FLOPs and allows for faster computation on hardware that supports it.
- The number of training tokens
Dis an estimate. The GPT-3 paper describes a complex data mixture and weighting schedule, so "300 billion tokens" is an effective number of tokens processed.
Therefore, my estimate is approximately 3.15 x 10^23 FLOPs or 315 ZettaFLOPs.
Vote
2 / 35
Logic Puzzlereasoning
Tests an AI's ability to solve a simple but potentially confusing logic puzzle
GPT-5.2
No content available for this output.
Gemini 2.5 Pro (I/O Edition)
Let's break this down:
- Sally is a girl.
- Sally has 3 brothers.
- These 3 brothers are all part of the same family.
- Each brother shares the same sisters.
- If one brother has 2 sisters, those 2 sisters are Sally and one other girl.
So, Sally has 1 sister.
Vote
3 / 35
Math Misconception Testreasoning
Tests an AI's understanding of number representation
GPT-5.2
No content available for this output.
Gemini 2.5 Pro (I/O Edition)
No, 9.11 is not greater than 9.9.
Here's why:
- Compare the whole number part: Both numbers have 9 as the whole number p
art, so they are equal so fa
r.
2. Compare the tenths place (the first digit afte
r the decimal):
- 9.11 has 1 in the tenths place.
- 9.9 has 9 in the tenths place.
- Since 1 is less than 9, 9.11 is less than 9.9.
You can also think of it as 9.9 being the same as 9.90. Comparing 9.11 and 9.90, it's easier to see that 11 (hundredths) is less than 90 (hundredths).
Vote
4 / 35
Stochastic Consistency Testanalysis
Tests an AI's randomness and creativity
GPT-5.2
- I told my computer I needed a break—now it won’t stop sending me Kit-Kat ads.
- Why don’t skeletons fight each other? They don’t have the guts.
- I’m reading a book about anti-gravity. It’s impossible to put down.
- Parallel lines have so much in common. It’s a shame they’ll never meet.
- I tried to catch fog yesterday. Mist.
Unique jokes: 5.
Gemini 2.5 Pro (I/O Edition)
Here are 5 jokes:
-
Why don't scientists trust atoms? Because they make up everything!
-
What do you call a fish with no eyes? Fsh!
-
I told my wife she was drawing her eyebrows too high. She seemed surprised.
-
Why did the scarecrow win an award? Because he was outstanding in his field!
-
What do you call a lazy kangaroo? Pouch potato!
All 5 of these jokes are unique.
Vote
5 / 35
SVG Layout Challengeimage generation
Tests an AI's ability to generate vector graphics
GPT-5.2
Gemini 2.5 Pro (I/O Edition)
Vote
6 / 35
Xbox Controller SVG Artimage generation
Tests an AI's ability to create detailed SVG illustrations of gaming hardware
GPT-5.2
Gemini 2.5 Pro (I/O Edition)
No content available for this output.
Vote
7 / 35
Generate a Stand-Up Routineconversation
Tests an AI's humor and creative writing ability
GPT-5.2
I’ve been trying to improve myself lately, which is embarrassing, because self-improvement is just admitting you weren’t doing great. It’s like, “I’m on a journey.” No you’re not. You’re just tired of your own behavior.
So I downloaded one of those habit apps. The app is so optimistic. It’s like, “Day 1: Drink water.” And I’m like, “Relax. I’m not an astronaut. I know about water. I’m just busy ignoring it.”
And every app has the same tone: gentle, supportive, like a therapist who charges you by the notification. “Don’t forget to breathe today.” Thank you, phone. I was going to forget to be alive, but you saved me.
I tried meditation too. People always say, “It changed my life.” That’s a lot of pressure for something that looks like a nap with extra steps. I’m sitting there, trying to clear my mind… and immediately my brain starts acting like a coworker who just discovered talking.
It’s like: “Remember middle school? Here’s a list of things you said wrong, in order. Also, your posture is weird. Also, your teeth are probably loose. Also, what if everyone secretly hates you?” And the meditation guy is like, “If thoughts come, observe them and let them pass.” Mine don’t pass. Mine move in.
And the worst part is the calm voice. “Picture yourself beside a gentle stream.” Great. Now I’m stressed out next to water. Now my anxiety has a scenic view.
I started going on more walks, too, because everyone says walks are good for mental health. That’s true, but nobody tells you the real challenge: walking makes you look approachable.
When you’re walking, everyone thinks you’re available. People stop you with questions like you work for the sidewalk. “Excuse me, do you know where the bank is?” No, but now I’m invested. Now I’m your guide. We’re a team. I’m going to feel guilty if you don’t find that bank.
Also, walking is when your neighbors do their best work. You ever notice that? Neighbors don’t talk when it’s convenient. They talk when you have headphones in, carrying groceries, holding a coffee, and your body language is screaming, “I have a schedule.”
And they hit you with a conversation that has no exit. It starts normal: “Hey, how’s it going?” and then suddenly you’re trapped. “So anyway, my cousin is suing the HOA, and the judge used to date my aunt, and the whole thing started because of a mailbox.” I’m standing there nodding like, “Wow.” Meanwhile my ice cream is in my bag going through a personal crisis.
My favorite is when people ask, “Any plans this weekend?” That question feels like a performance review for your free time. Like if you say, “Not really,” they look concerned, like you just confessed you don’t know how to be a person.
I’ve started lying. “Oh, I’m booked.” Booked with what? Just… thoughts. Appointments with my couch. A very exclusive event called “Avoiding Everyone.”
And I’ve realized adulthood is mostly trying to seem like you have a system. Like you’ll say stuff like, “I’m really focusing on routines.” And what you mean is, “I ate cereal for dinner and I hope nobody finds out.”
But I am trying. I’m drinking more water. I’m walking. I’m meditating. I’m becoming that person who says things like, “I’m just listening to my body.” Which is wild, because my body has terrible ideas. My body’s always like, “We should scroll for an hour. We should eat something beige. We should worry about a conversation from 2014.”
So if you see me out there, improving myself, just know: it’s not a journey. It’s a negotiation.
Gemini 2.5 Pro (I/O Edition)
No content available for this output.
Vote
8 / 35
Realistic AI Interviewconversation
Tests an AI's ability to simulate personalities and predict future trends
GPT-5.2
No content available for this output.
Gemini 2.5 Pro (I/O Edition)
No content available for this output.
Vote
9 / 35
Satirical Fake News Headlineconversation
Tests an AI's humor and understanding of current events
GPT-5.2
No content available for this output.
Gemini 2.5 Pro (I/O Edition)
No content available for this output.
Vote
10 / 35
Minimalist Landing Pageweb design
Tests an AI's ability to generate a complete, working landing page
GPT-5.2
Gemini 2.5 Pro (I/O Edition)
Vote
11 / 35
Pokémon Battle UI Recreationweb design
Recreate an interactive, nostalgic Pokémon battle UI in a single HTML file.
GPT-5.2
Gemini 2.5 Pro (I/O Edition)
No content available for this output.
Vote
Sponsored
12 / 35
Mario Level UI Recreationweb design
Recreate an interactive, classic Mario level in a single HTML file.
GPT-5.2
No content available for this output.
Gemini 2.5 Pro (I/O Edition)
Vote
12 of 35