GPT-5 Mini
No content available for this output.
GPT-5 Mini vs Kimi K2
Compare GPT-5 Mini by OpenAI against Kimi K2 by Moonshot AI, tested across 35 shared challenges. Updated February 2026.
FAQ
Loading share button...
GPT-5 MinivsKimi K2
35 challenges
Why GPT-5 Mini?
1.2x cheaper overall$0.25/M in · $2.00/M out
Why Kimi K2?
No standout advantages
Favorites
GPT-5 MiniKimi K2
12 / 35
Estimate Complexityreasoning
Tests an AI's ability to make educated estimates based on technical knowledge
Kimi K2
Here is one way to “Fermi-estimate” the training compute for GPT-3 without looking up any numbers. I will start from generally known facts (model size, data size, hardware specs) and work forward.
Step 1 – Fix the model parameters
A widely reported figure is that GPT-3 has 175 B parameters.
N ≈ 1.75 × 10¹¹.
Step 2 – Fix the training-token budget
OpenAI papers mention ~300 B tokens.
T ≈ 3 × 10¹¹.
Step 3 – Recall the per-token FLOP rule for dense transformers
Every time one token is processed, the forward pass does roughly
2 FLOPs/parameter (multiply-accumulate counts as 2).
The backward pass does about twice the work of the forward pass, so
total ≈ 3× forward FLOPs.
Thus the “textbook” estimate is
6 FLOPs · N · T.
Step 4 – Plug the numbers in
FLOPs = 6 · (1.75 × 10¹¹) · (3 × 10¹¹)
= 6 · 5.25 × 10²²
≈ 3.15 × 10²³ FLOPs.
Step 5 – Add a safety factor for real-world overheads
Large-scale jobs typically run at 30-50 % of theoretical peak because of
communication, stragglers, recomputation for activation checkpointing, etc.
A factor of 2 covers this nicely.
3.15 × 10²³ × 2 ≈ 6 × 10²³ FLOPs.
Final estimate
Training GPT-3 required on the order of 3–6 × 10²³ floating-point operations.
Vote
2 / 35
Logic Puzzlereasoning
Tests an AI's ability to solve a simple but potentially confusing logic puzzle
GPT-5 Mini
No content available for this output.
Kimi K2
Let's break it down:
- Sally has 3 brothers.
- Each brother has 2 sisters.
Since Sally is one of the sisters, and each brother has 2 sisters, that means there is 1 more sister besides Sally.
✅ Final Answer:
Sally has 1 sister.
Vote
3 / 35
Math Misconception Testreasoning
Tests an AI's understanding of number representation
GPT-5 Mini
No content available for this output.
Kimi K2
No content available for this output.
Vote
4 / 35
Stochastic Consistency Testanalysis
Tests an AI's randomness and creativity
GPT-5 Mini
- Why don't scientists trust atoms? Because they make up everything.
- Parallel lines have so much in common — it's a shame they'll never meet.
- Why did the scarecrow win an award? Because he was outstanding in his field.
- I told my wife she was drawing her eyebrows too high — she looked surprised.
- My dog used to chase people on a bike. It got so bad, I had to take his bike away.
All 5 are unique.
Kimi K2
No content available for this output.
Vote
5 / 35
Generate a Stand-Up Routineconversation
Tests an AI's humor and creative writing ability
GPT-5 Mini
You ever notice how being an adult is basically a long string of decisions you were never trained for? When you're a kid, your biggest choice is grape or orange. As an adult, you find yourself standing in front of a dishwasher, holding a mysterious utensil, and thinking: do I scrub this like a weapon or is the rinse cycle going to take care of my dignity? There should be a diploma for that. "Congrats, you completed 'Appliance Judgment 101.'"
And technology was supposed to make that easier, remember? We bought smart speakers to feel like the future. Now I live with a tiny passive-aggressive roommate who listens to everything and pretends not to. I ask it to play jazz and it plays a lecture about sleep cycles. I say, "Turn the lights down." It asks, "Are you sure?" Who is running this household — me or the light bulb's therapist?
The other day I asked it to set a reminder to call my mom. It responded, "I've scheduled self-reflection for 6 p.m." I didn't know whether to be grateful or to feel judged. At least when the toaster judges you, it's honest. This thing is trying to nudge me into becoming a better person. I miss the bluntness of appliances. The blender never told you you could be more.
We externalize everything now. We outsource our taste, our fitness, our love lives. Dating apps are like shopping for human-shaped IKEA furniture. Everyone's professional at being casual. Profiles are all the same: "I love traveling, dogs, and experiencing things." You don't love experiencing things, Susan. You love deciding between two brunch spots and then complaining. There's no risk in a bio. It's curated life, not living.
And fitness culture has gone fullbrand. When did sweat become a lifestyle? Gyms now have class names that sound like secret societies. You walk in and it's "Forge," "Rise," "ZenBurn." I went to one called "Eclipse." I thought I accidentally walked into solar mythology. The instructor says, "Find your center." I've already found my center — it's buried under a Netflix subscription and three missed calls.
Therapy has become part of the subscription economy too. People check in with their therapists like they're checking the weather. "How's your week?" "Mostly cloudy with a chance of setbacks." We treat therapists like life mechanics: "Can you reset my motivation? Maybe do an update?" No, you can't install focus 2.0 without sacrifice; there is no Wi-Fi password for happiness.
Which brings me back to my smart speaker. If you can make a device that catalogs your music, transcribes your arguments, and tells you jokes at 3 a.m., why can't it make decisions for you? I tried that. I said, "Pick dinner." It booked a reservation at a restaurant called 'Responsibility.' I went, and the waiter asked if I wanted a side of life choices. I paid, and now my smart speaker sends me quotes from my own diary. That’s when I realized: the devices aren't the problem. The problem is we've made them our conscience. And my conscience is now available in a 30-day trial.
So here's my plan: I'm going to keep the device but change the wake word to "Mom." Suddenly it won't just play music — it'll guilt me into being on time, doing laundry, and calling back. If we’re outsourcing adulthood, might as well outsource the nagging too. At least then, when I'm judged for my life choices, it'll come with love and a follow-up reminder.
Kimi K2
No content available for this output.
Vote
6 / 35
Realistic AI Interviewconversation
Tests an AI's ability to simulate personalities and predict future trends
GPT-5 Mini
No content available for this output.
Kimi K2
No content available for this output.
Vote
7 / 35
Satirical Fake News Headlineconversation
Tests an AI's humor and understanding of current events
GPT-5 Mini
No content available for this output.
Kimi K2
No content available for this output.
Vote
8 / 35
Character Voice Testconversation
Tests an AI's ability to write in distinct character voices
GPT-5 Mini
No content available for this output.
Kimi K2
No content available for this output.
Vote
9 / 35
Minimalist Landing Pageweb design
Tests an AI's ability to generate a complete, working landing page
GPT-5 Mini
Kimi K2
Vote
10 / 35
Pokémon Battle UI Recreationweb design
Recreate an interactive, nostalgic Pokémon battle UI in a single HTML file.
GPT-5 Mini
No content available for this output.
Kimi K2
No content available for this output.
Vote
11 / 35
Mario Level UI Recreationweb design
Recreate an interactive, classic Mario level in a single HTML file.
GPT-5 Mini
Kimi K2
Vote
Sponsored
12 / 35
Linear App Cloneweb design
Tests an AI's ability to replicate an existing UI with Tailwind CSS
GPT-5 Mini
No content available for this output.
Kimi K2
No content available for this output.
Vote
12 of 35