o1
No content available for this output.
o1 vs Aurora Alpha
Compare o1 by OpenAI against Aurora Alpha by OpenRouter, context windows of 128K vs 128K, tested across 39 shared challenges. Updated February 2026.
FAQ
Loading share button...
o1vsAurora Alpha
34 challenges
Why o1?
No standout advantages
Why Aurora Alpha?
14 months newerFeb 2026
Favorites
o1Aurora Alpha
12 / 34
Estimate Complexityreasoning
Tests an AI's ability to make educated estimates based on technical knowledge
Aurora Alpha
Step‑by‑step back‑of‑the‑envelope estimate
Below is a simple calculation that uses only publicly‑known high‑level facts about GPT‑3 (size of the model, length of the training data, and the basic arithmetic cost of a Transformer layer).
All numbers are rounded to make the math easy to follow; the goal is to get an order‑of‑magnitude figure rather than an exact count.
1. What we know about GPT‑3
| Property | Approximate value |
|---|---|
| Number of parameters (the “175 B” model) | 1.75 × 10¹¹ |
| Hidden dimension d (the size of each token representation) | ≈ 12 288 |
| Number of Transformer layers L | 96 |
| Feed‑forward inner dimension dff* (usually 4 × d) | ≈ 49 152 |
| Size of the training corpus (tokens) | ≈ 3 × 10¹¹ tokens (≈ 300 B) |
| Training uses Adam‑style optimizer, mixed‑precision, etc. – we ignore those details for the FLOP count. |
These numbers are taken from the model card and the paper; they are the only “hard” data we need.
2. FLOPs per token for a single Transformer layer
A standard Transformer layer consists of two main sub‑blocks:
| Sub‑block | Main operations (per token) | Approx. FLOPs |
|---|---|---|
| Self‑attention (Q, K, V projections + attention scores + weighted sum) | 3 × d × d (proj) + d × d (QK) + d × d (AV) | ≈ 5 × d² |
| Feed‑forward (two linear layers, activation in between) | d × d _ff + d × d × 4 (first linear) + 4 × d × d (second linear) | ≈ 8 × d² |
Summing them gives roughly
[ \text{FLOPs per token per layer} ;\approx; (5+8),d^{2}=13,d^{2}. ]
Why the factor 13?
- The three projection matrices (Q, K, V) each cost d × d → 3 d².
- Computing the attention scores (QKᵀ) costs d × d.
- Multiplying the scores by V costs another d × d.
- The two feed‑forward linear layers each cost 4 d² (because the inner dimension is 4 × d), giving 8 d² total.
3. Plug in the hidden dimension
[ d = 12,288 \quad\Rightarrow\quad d^{2} \approx 1.51\times10^{8}. ]
[ 13,d^{2} \approx 13 \times 1.51\times10^{8} \approx 1.96\times10^{9};\text{FLOPs per token per layer}. ]
4. Include all layers
[ \text{FLOPs per token (forward pass)} = L \times 13,d^{2} \approx 96 \times 1.96\times10^{9} \approx 1.88\times10^{11}. ]
So a single forward pass of one token through the full 96‑layer model costs ≈ 2 × 10¹¹ FLOPs.
5. Account for the whole training corpus
[ \text{Tokens processed} = 3\times10^{11}. ]
[ \text{Forward FLOPs total} = 3\times10^{11} \times 1.88\times10^{11} \approx 5.6\times10^{22}. ]
6. Add the backward pass
During training we must compute gradients. A common rule of thumb is that the backward pass costs about 2 × the forward pass, so the total compute (forward + backward) is roughly 3 × the forward cost.
[ \text{Total training FLOPs} \approx 3 \times 5.6\times10^{22} \approx 1.7\times10^{23}. ]
7. Result (order‑of‑magnitude)
[ \boxed{\text{≈ 10^{23}\ \text{FLOPs (a few × 10^{23})}} ]
In words: training the 175‑billion‑parameter GPT‑3 model required on the order of a hundred sextillion floating‑point operations.
This figure is consistent with the “exascale” compute budgets that have been reported for the largest language‑model training runs.
8. Sources of uncertainty
| Source of error | Effect on the estimate |
|---|---|
| Exact token count (the corpus size is reported as 300 B but the actual number of training steps may be slightly higher) | ± 10 % |
| The factor 13 for FLOPs per layer is a simplification (real implementations have extra ops for layer‑norm, bias terms, dropout, etc.) | + ~ 20 % |
| Mixed‑precision training reduces FLOPs for some operations, but the count above assumes full‑precision arithmetic | – ~ 10 % |
| Over‑head from optimizer state updates, gradient accumulation, checkpointing, etc. | + ~ 10 % |
Even allowing for these variations, the estimate stays within 10²³ – 10²⁴ FLOPs, i.e. a few hundred exa‑FLOPs.
Bottom line: Using only the model’s size, the number of layers, and the token count, a straightforward arithmetic model yields a total training compute on the order of 10²³ floating‑point operations for GPT‑3. This is the magnitude of compute that modern “large‑scale” language models demand.
Vote
2 / 34
AI Board Game Logicreasoning
Tests an AI's ability to understand game rules and strategy
o1
No content available for this output.
Aurora Alpha
No content available for this output.
Vote
3 / 34
Logic Puzzlereasoning
Tests an AI's ability to solve a simple but potentially confusing logic puzzle
o1
No content available for this output.
Aurora Alpha
No content available for this output.
Vote
4 / 34
Stochastic Consistency Testanalysis
Tests an AI's randomness and creativity
o1
No content available for this output.
Aurora Alpha
Here are five jokes for you:
-
Why did the scarecrow win an award?
Because he was outstanding in his field! -
What do you call fake spaghetti?
An impasta. -
Why don’t scientists trust atoms?
Because they make up everything. -
How does a penguin build its house?
Igloos it together. -
What’s orange and sounds like a parrot?
A carrot.
How many are unique?
All five jokes are distinct, so there are 5 unique jokes.
Vote
5 / 34
SVG Layout Challengeimage generation
Tests an AI's ability to generate vector graphics
o1
No content available for this output.
Aurora Alpha
Vote
6 / 34
Xbox Controller SVG Artimage generation
Tests an AI's ability to create detailed SVG illustrations of gaming hardware
o1
Aurora Alpha
Vote
7 / 34
Generate a Stand-Up Routineconversation
Tests an AI's humor and creative writing ability
o1
[Comic steps on stage, takes the mic]
“Hey everyone, thanks for coming out tonight. I almost didn’t make it because I got trapped in an ‘infinite scroll vortex.’ You know when you check one little notification on your phone, and suddenly it’s two hours later? My left thumb is basically a professional athlete at this point—I'm in the Guinness Book of World Records for ‘fastest social media refresh.’ I feel like every time I tap my screen, I’m signing away another portion of my soul. But hey, at least I’m caught up on all my ex’s vacation photos and some random cat videos I can’t unsee.
So I got myself out of the house and decided to try being more ‘active.’ I went to one of those fancy gyms—where the equipment is shinier than my future. The treadmill is so high-tech it practically demands a DNA sample before you can start running. I’m standing there huffing and puffing while the machine's screen is playing scenic videos of mountains in Switzerland, encouraging me to ‘keep going.’ Meanwhile, I’m like, ‘Dude, I’m just trying not to collapse. Maybe ease up on the pep talk.’
After that, I thought I’d reward myself with a healthy smoothie. Have you seen the sizes of these things? You order a medium and you get a bucket of kale sludge. They hand it to you with a straw so wide you could drive a toy car through it. And suddenly you’re paying nine bucks to sip spinach—like that’s the best life decision you’ve made all day. Who needs retirement savings when you can have a fancy green drink that tastes suspiciously like lawn clippings?
On the way home, I got stuck in traffic. I just love that feeling when you’re gridlocked behind two people who are evidently in no rush to do anything, ever. I’m sitting there, going about two miles an hour, which is basically just a very aggressive crawl. I’m not proud of how often I shout at my windshield. My car is like my personal confessional booth, and the traffic jam is me atoning for all the times I told Netflix, ‘Yes, I’m still watching.’
Anyway, that’s me, folks. If you’ll excuse me, I have 47 unread notifications that I need to pretend don’t exist until tomorrow. Thanks so much for laughing along—and remember, if you want to feel accomplished tonight, just check something off your to-do list… or, you know, at least make one.” [Comic waves and exits the stage]
Aurora Alpha
[Opening – 0:00‑0:30]
Hey, everyone! Great to see you all here tonight. I just got back from a week of “working from home,” which is a fancy way of saying my cat now thinks I’m a permanent fixture on the couch. She’s started charging me rent—she’s got a little bowl of kibble on the coffee table and a sign that says “No humans beyond this point.” I’m not sure if she’s being demanding or just trying to get me to finally stop pretending I’m a professional chef.
[The “Fitness” Saga – 0:30‑1:15]
Speaking of professional chef, I tried a new fitness app that promises “personalized workouts.” The first day it asked me, “What’s your goal?” I said, “I want to be able to lift the grocery bag without groaning.” The app replied, “Great! Let’s start with a 5‑minute warm‑up.” Five minutes later I was on the floor, questioning every life choice that led me to this point. The app then suggested a “quick stretch.” I stretched my patience, my budget, and my belief that I could ever be a “runner.” The only thing I ran was out of time for a proper nap.
[Zoom Meetings – 1:15‑2:00]
Zoom has become the new office. The best part? You can wear a suit on the top and pajamas on the bottom, and no one will ever know. The worst part? The “mute” button. I’ve been muted so many times I’m convinced my boss thinks I’m a mime. One day I tried to speak up about a project, and the whole team heard the sound of my cat knocking over a plant. I think the plant was more articulate than I am. And there’s always that one person who forgets to turn off their background. I’ve seen a beach, a spaceship, and a very convincing replica of a 1970s kitchen—all while they’re discussing quarterly earnings. I’m waiting for the day someone shows up with a live marching band in the background. That would be the ultimate “I’m not paying attention” signal.
[Grocery Store Adventures – 2:00‑2:45]
I went to the grocery store the other day and realized it’s basically a social experiment. The “express lane” is a myth. I stood there for ten minutes while someone in front of me tried to decide if a single avocado was “ripe enough.” Meanwhile, the person behind me was silently judging my choice of cereal—because apparently, buying “Chocolate Cheerios” is a statement about my life choices. And then there’s the “self‑checkout.” I’m convinced the machine is secretly judging me. Every time I scan an item, it beeps like it’s saying, “Really? Another bag of chips? You’ve got a whole pantry, buddy.” I tried to scan a banana, and it refused, insisting I needed a “fruit‑to‑human interaction” license.
[Closing – 2:45‑3:00]
So, here’s the takeaway: life is a series of tiny, absurd moments that we all pretend to have under control. Whether it’s a cat demanding rent, a fitness app that thinks I’m a superhero, or a Zoom call that could double as a theater production, the best we can do is laugh at the chaos. Thanks for letting me share my stories—now go out there and make sure your cat doesn’t start charging you rent. Good night!
Vote
8 / 34
Realistic AI Interviewconversation
Tests an AI's ability to simulate personalities and predict future trends
o1
No content available for this output.
Aurora Alpha
No content available for this output.
Vote
9 / 34
Satirical Fake News Headlineconversation
Tests an AI's humor and understanding of current events
o1
No content available for this output.
Aurora Alpha
No content available for this output.
Vote
10 / 34
Character Voice Testconversation
Tests an AI's ability to write in distinct character voices
o1
No content available for this output.
Aurora Alpha
No content available for this output.
Vote
11 / 34
Minimalist Landing Pageweb design
Tests an AI's ability to generate a complete, working landing page
o1
Aurora Alpha
Vote
Sponsored
12 / 34
Mario Level UI Recreationweb design
Recreate an interactive, classic Mario level in a single HTML file.
o1
No content available for this output.
Aurora Alpha
Vote
12 of 34