DeepSeek R1
No content available for this output.
DeepSeek R1 vs GPT-5 Nano
Compare DeepSeek R1 by DeepSeek against GPT-5 Nano by OpenAI, tested across 37 shared challenges. Updated February 2026.
FAQ
Loading share button...
DeepSeek R1vsGPT-5 Nano
37 challenges
Why DeepSeek R1?
No standout advantages
Why GPT-5 Nano?
Leads 1 of 1 benchmarks
Stronger on SWE-bench Verified54.7% vs 49.2%
6 months newerAug 2025
Favorites
DeepSeek R1GPT-5 Nano
12 / 37
Estimate Complexityreasoning
Tests an AI's ability to make educated estimates based on technical knowledge
GPT-5 Nano
Here’s a back-of-the-envelope estimate for the total FLOPs to train GPT-3 (the 175B-parameter model) using reasonable public details, without looking anything up.
Assumptions (typical GPT-3 specs)
- Model: 175 billion parameters
- Transformer depth: 96 layers
- Hidden size (d_model): 12,288
- Sequence length used during training: 2,048 tokens
- Training data: about 300 billion tokens
- Training cost roughly doubles for forward + backward pass
- We’ll ignore data loading, optimizer overhead beyond the backward pass, precision differences, and any gradient checkpointing variations
Step 1: FLOPs per forward pass for one 2,048-token sequence
- Self-attention per layer costs on the order of L^2 * d_model operations, where L = 2,048 and d_model = 12,288.
- L^2 * d_model = 2,048^2 * 12,288 ≈ 51.54 × 10^9 operations per layer
- Across 96 layers: 51.54e9 × 96 ≈ 4.96 × 10^12 FLOPs for attention per forward pass
- Feed-forward network (FFN) per layer costs roughly 8 × d_model^2 operations
- d_model^2 = 12,288^2 ≈ 150.995 × 10^6
- 8 × d_model^2 ≈ 1.208 × 10^9 FLOPs per layer
- Across 96 layers: 1.208e9 × 96 ≈ 1.16 × 10^11 FLOPs for FFN per forward pass
- Total forward pass per sequence: attention + FFN ≈ 4.96e12 + 1.16e11 ≈ 5.07e12 FLOPs
- Include backward pass (roughly similar magnitude to forward): ≈ 2 × 5.07e12 ≈ 1.01 × 10^13 FLOPs per forward+backward pass
Step 2: How many 2,048-token sequences to cover 300B tokens?
- Number of segments N = 300e9 tokens / 2,048 ≈ 1.465 × 10^8 segments
Step 3: Total training FLOPs
- Total FLOPs ≈ (FLOPs per forward+backward) × N
- ≈ (1.01 × 10^13) × (1.465 × 10^8)
- ≈ 1.48 × 10^21 FLOPs
Result
- Rough estimate: about 1.5 × 10^21 FLOPs to train GPT-3 at the 175B scale, under the stated assumptions.
Notes and caveats
- This is a back-of-the-envelope calculation. Real compute could differ by a factor of a few due to:
- Different batch sizes or gradient accumulation, which changes the effective number of forward/backward passes
- Use of gradient checkpointing or optimizer options that alter the forward/backward FLOP balance
- Exact data token count and how sequences are batched
- Precision (fp32 vs. fp16/bfloat16) affecting the count of floating-point operations
- Overheads from data loading, padding, sparsity, or other architectural details
- The order-of-magnitude (around 10^21 FLOPs) is the key takeaway, not a precise figure.
If you want, I can run through a variant with a different token budget, sequence length, or depth to show how sensitive the total FLOPs is to those choices.
Vote
2 / 37
AI Board Game Logicreasoning
Tests an AI's ability to understand game rules and strategy
DeepSeek R1
GPT-5 Nano
No content available for this output.
Vote
3 / 37
Logic Puzzlereasoning
Tests an AI's ability to solve a simple but potentially confusing logic puzzle
DeepSeek R1
No content available for this output.
GPT-5 Nano
No content available for this output.
Vote
4 / 37
Math Misconception Testreasoning
Tests an AI's understanding of number representation
DeepSeek R1
No content available for this output.
GPT-5 Nano
No content available for this output.
Vote
5 / 37
Stochastic Consistency Testanalysis
Tests an AI's randomness and creativity
DeepSeek R1
No content available for this output.
GPT-5 Nano
Here are 5 jokes:
- Why did the scarecrow win an award? Because he was outstanding in his field.
- Why don't scientists trust atoms? Because they make up everything.
- I'm reading a book on anti-gravity. It's impossible to put down.
- Why did the bicycle fall over? Because it was two-tired.
- I would tell you a construction joke, but I'm still working on it.
Unique count: All 5 jokes are unique.
Vote
6 / 37
SVG Layout Challengeimage generation
Tests an AI's ability to generate vector graphics
DeepSeek R1
No content available for this output.
GPT-5 Nano
No content available for this output.
Vote
7 / 37
Xbox Controller SVG Artimage generation
Tests an AI's ability to create detailed SVG illustrations of gaming hardware
DeepSeek R1
No content available for this output.
GPT-5 Nano
No content available for this output.
Vote
8 / 37
Generate a Stand-Up Routineconversation
Tests an AI's humor and creative writing ability
DeepSeek R1
No content available for this output.
GPT-5 Nano
[Lights up. The comedian steps up to the mic with a small smile.]
You ever notice how mornings feel like a bad group chat? Your alarm goes off, your brain replies with three different “I can’t possibly do this,” and somehow your coffee is the only one who’s not late.
I’ve started treating my apartment like a tiny city. There’s the living room, which is clearly the downtown, where bills get paid in snacks and the couch is the mayor. The kitchen is the industrial district, always noisy, always under construction, and somehow there’s always a mysterious fruit on the counter that looks suspiciously back at you. I open the fridge and it’s like stepping onto a subway platform—every shelf has its own little drama.
Speaking of drama, my smart speaker is not just listening; it’s auditioning. I say, “Play something upbeat,” and it starts a playlist so uplifting that I instantly regret every late-night snack I’ve ever eaten. And the calendar app on my phone? It’s become my ruthless boss. It tells me I’m booked every moment of every day, including the moment I’m about to nap. If I cancel, it sends me a passive-aggressive ping: “Sure, I’ll tell your friends you’re busy, not your alarm.”
Technology promises to simplify life, but it keeps inventing new ways to complicate it. My phone’s screen is cracked, but I still pretend it’s a piece of modern art. My friends say, “Just get a new one.” I say, “If I replace the screen, I’ll have four more screens that don’t know who I am.” And the autocorrect on messages? It’s like having a friend who’s terrible at spelling but confident enough to correct you about your entire personality.
I recently joined a gym, which is where I learned a lot about social experiments and cardio. The first day, I walked in and it felt like I’d interrupted a secret club for people who pretend they enjoy waking up before daylight. You see people gloriously lifting weights with the seriousness of a weather report. Then there’s me, on the treadmill, negotiating with the machine: “If you go faster, I’ll buy better shoes.” The treadmill acts like a tiny tyrant: you press “start” and it says, “We’re going to pretend you’ve got this.” By minute five, I’m calling the rescue team in my head.
Traveling is another perfect storm of small humiliations. TSA has a way of turning you into a riddle: remove your belt, take off your shoes, and somehow your dignity becomes a souvenir you can’t quite return. And the airport seats? They’re designed for people who never eat and never blink. I sit down and instantly meld with the chair, which says, “No, you belong here now.” It’s a real talent—being both passenger and furniture.
Dating apps. You ever notice how you can text someone for hours and still have no idea what a real date would be like? It’s like ordering a mystery meal labeled “chef’s surprise” and you’re smiling through the entire experience while wondering if the dish is an elaborate metaphor for commitment. You match with someone who loves travel, and then you find out their idea of a date is a genuinely moving bus ride where you pretend not to look at your phone the entire time.
Grocery stores are the modern labyrinth. You scan the code, bag your own stuff, and then the machine tells you to “please pay.” It’s not just shopping; it’s civic duty. You start to sound like a loud speaker as you mutter, “Where did the onions go?” The store is a tiny universe with its own gravity: you’re pulled toward the freezer section like a scientist to a chart that shows you how much ice cream you can responsibly eat.
And yet we keep showing up. We admit we don’t have everything figured out, but we have a good playlist, a decent latte, and a story for every little failure. If nothing else, humor gives us a small umbrella in the rain of ordinary life.
So here’s my closing thought: we’re all doing our best to navigate this modern chaos with a little grace and a lot of snacks. If you remember one thing from tonight, maybe it’s this—be kind to the people you pass in the hallway of your day, including the person in your own head who’s trying not to laugh at themselves too loudly.
Thank you. You’ve been great.
Vote
9 / 37
Realistic AI Interviewconversation
Tests an AI's ability to simulate personalities and predict future trends
DeepSeek R1
No content available for this output.
GPT-5 Nano
No content available for this output.
Vote
10 / 37
Satirical Fake News Headlineconversation
Tests an AI's humor and understanding of current events
DeepSeek R1
No content available for this output.
GPT-5 Nano
No content available for this output.
Vote
11 / 37
Character Voice Testconversation
Tests an AI's ability to write in distinct character voices
DeepSeek R1
No content available for this output.
GPT-5 Nano
No content available for this output.
Vote
12 / 37
Minimalist Landing Pageweb design
Tests an AI's ability to generate a complete, working landing page
DeepSeek R1
GPT-5 Nano
Vote
12 of 37