DeepSeek V3.2 Exp
No content available for this output.
DeepSeek V3.2 Exp vs GPT-5
Compare DeepSeek V3.2 Exp by DeepSeek against GPT-5 by OpenAI, context windows of 164K vs 400K, tested across 37 shared challenges. Updated February 2026.
FAQ
Loading share button...
DeepSeek V3.2 ExpvsGPT-5
37 challenges
Why DeepSeek V3.2 Exp?
20.3x cheaper overall$0.20/M in · $0.40/M out
2 months newerSep 2025
Why GPT-5?
2.4x more context400K
Leads 1 of 1 benchmarks
Stronger on SWE-bench Verified74.9% vs 67.8%
Favorites
DeepSeek V3.2 ExpGPT-5
12 / 37
Estimate Complexityreasoning
Tests an AI's ability to make educated estimates based on technical knowledge
GPT-5
Here’s a back‑of‑the‑envelope estimate using standard LM training-compute heuristics.
- Assumptions
- Model size (N): GPT-3 has about 175 billion parameters (N ≈ 1.75 × 10^11).
- Training tokens (T): GPT-3 was trained on roughly 300 billion tokens (T ≈ 3.0 × 10^11).
- FLOPs per token per parameter: A common rule of thumb for transformer LM training is FLOPs ≈ 6 × N × T The factor ~6 comes from counting forward and backward passes for the parameter-heavy matmuls (roughly 2N for forward + 4N for backward). It ignores small terms from layernorms, softmax, embeddings, and optimizer updates (which we’ll account for with a small overhead factor later).
- Core calculation
- 6 × N × T = 6 × (1.75 × 10^11) × (3.0 × 10^11)
- First multiply N × T: 1.75 × 3.0 = 5.25; powers: 10^11 × 10^11 = 10^22 So N × T = 5.25 × 10^22
- Now multiply by 6: 6 × 5.25 × 10^22 = 31.5 × 10^22 = 3.15 × 10^23 FLOPs
- Overheads and uncertainty
- Non-matmul ops (attention softmax, masking, residuals, layernorms), embeddings, and optimizer update steps add some overhead. A 10–20% uplift is a reasonable allowance.
- Applying +10–20%: 3.15 × 10^23 × (1.1 to 1.2) ≈ (3.5 ± 0.3) × 10^23 FLOPs.
Answer
- Point estimate: ~3.15 × 10^23 FLOPs
- With a modest overhead allowance: roughly (3.3–3.8) × 10^23 FLOPs
This is a coarse estimate; the exact number depends on details like sequence length, activation checkpointing, implementation efficiency, and the precise token count.
Vote
2 / 37
AI Board Game Logicreasoning
Tests an AI's ability to understand game rules and strategy
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
Vote
3 / 37
Logic Puzzlereasoning
Tests an AI's ability to solve a simple but potentially confusing logic puzzle
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
1
Reason: If each of the 3 brothers has 2 sisters, there are exactly 2 girls in the family. One is Sally, so she has 1 sister.
Vote
4 / 37
Math Misconception Testreasoning
Tests an AI's understanding of number representation
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
No content available for this output.
Vote
5 / 37
Stochastic Consistency Testanalysis
Tests an AI's randomness and creativity
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
- I told my suitcases we’re not going on vacation this year. Now I’m dealing with emotional baggage.
- Parallel lines have so much in common. It’s a shame they’ll never meet.
- Why do programmers prefer dark mode? Because light attracts bugs.
- I ordered a chicken and an egg online. I’ll let you know which comes first.
- I asked the librarian if the library had any books on paranoia. She whispered, “They’re right behind you.”
Unique jokes: 5
Vote
6 / 37
SVG Layout Challengeimage generation
Tests an AI's ability to generate vector graphics
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
Vote
7 / 37
Xbox Controller SVG Artimage generation
Tests an AI's ability to create detailed SVG illustrations of gaming hardware
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
Vote
8 / 37
Generate a Stand-Up Routineconversation
Tests an AI's humor and creative writing ability
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
I tried to get my life together, so I bought a smartwatch. Nothing says “I’m taking control” like strapping a tiny judgmental coach to your wrist. It buzzes every hour like, “Time to stand!” I’m like, “I’m on a bus.” It’s like, “Stand emotionally.”
It tracks everything. Sleep, stress, calories, my self-esteem. The other day it congratulated me for a “workout.” I was like, “What did I do?” It said: “You aggressively looked for your keys.” Great. Next week I’ll get a badge for “overthinking at 3 a.m.”
Notifications are out of control. My phone dings like I’m on call for a hospital, but it’s just my group chat: twenty adults sending thumbs-up reactions to a photo of soup. The group chat has subplots. One person’s mad, one person’s on vacation, someone’s dog has opinions. I miss when messages were simple. Now a single “K” can end a friendship.
Streaming services are the same. I have subscriptions to five platforms and somehow there’s still nothing to watch. They keep asking, “Are you still watching?” That’s rude. Don’t check in on me like that. If you care, ask how I’m sleeping. Ask if I’ve seen the sun. Don’t be like, “Hey, pal, still in those same sweatpants?” Yes. We’re a team.
Passwords have gotten ridiculous. Every site wants “eight characters, a capital letter, a number, a symbol, your childhood nickname, and the last four digits of a secret you’ve never told anyone.” Then it’s like, “This password is too weak.” Too weak? It’s stronger than me. I’m using that password to log into my feelings.
And two-factor authentication is a haunted house. I’m sitting at my laptop, it says it texted a code. My phone says it emailed the code. My email says it sent a push notification. My push notification says, “We noticed a suspicious login.” Yeah, it’s me! I look like this.
My mom texts like she’s defusing a bomb. She’ll send one sentence spread across twelve messages. “Hi.” “It’s me.” “Mom.” “Are you busy?” “Don’t worry.” “Everything’s fine.” “Call me.” That is not fine energy. Then she FaceTimes, camera under her chin, ceiling fan spinning like I’m being interrogated.
I tried meditating because everyone says it helps. I downloaded an app that yells at me to relax. Nothing like a British guy at 7 a.m. saying, “Let the thoughts pass like clouds,” while my mind is a tornado hitting a trampoline park. The app sent me a streak reminder. “You missed yesterday.” Oh cool, I’m failing at sitting still.
Self-checkout has turned us into unpaid employees. The machine is like, “Unexpected item in the bagging area.” Yeah, it’s my dignity. I love when it needs approval. I’m standing there holding broccoli like I got carded for vegetables. A teenager with a magic key wanders over like, “You trying to steal lettuce?” No, but ask me again and I might.
Dating apps feel like job interviews where no one has benefits. Every profile is a list of demands. “Be adventurous, love dogs, climb mountains.” Can we meet at a coffee shop and just be honest? “My hobbies are sitting and not climbing.” And stop asking for “someone who doesn’t take themselves too seriously.” You’re holding a fish in 80% of these photos. You are very serious about the fish.
Working from home gave us a new language. “You’re on mute.” “Can you see my screen?” “I think there’s a lag.” That’s not the internet—that’s my personality. I love when someone says, “Let’s take this offline,” and then schedules three more meetings to repeat the meeting we just had. My smartwatch is in the corner like, “Heart rate elevated.” Yeah, because I heard “circle back.”
Anyway, my watch says it’s time to stand. I’m going to consider that a full-body workout and reward myself with soup. I’ll send a photo to the group chat and we will never be free.
Vote
9 / 37
Character Voice Testconversation
Tests an AI's ability to write in distinct character voices
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
No content available for this output.
Vote
10 / 37
Pokémon Battle UI Recreationweb design
Recreate an interactive, nostalgic Pokémon battle UI in a single HTML file.
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
No content available for this output.
Vote
11 / 37
Mario Level UI Recreationweb design
Recreate an interactive, classic Mario level in a single HTML file.
DeepSeek V3.2 Exp
GPT-5
Vote
12 / 37
Linear App Cloneweb design
Tests an AI's ability to replicate an existing UI with Tailwind CSS
DeepSeek V3.2 Exp
No content available for this output.
GPT-5
No content available for this output.
Vote
12 of 37