Updated Jul 11, 2025

Our Verdict

Kimi K2Winner

Claude 3.7 SonnetRunner-up

Pick Kimi K2. In 25 blind votes, Kimi K2 wins 75% of the time. That's not luck.

Kimi K2 particularly excels in Conversation, Reasoning, Web Design. Kimi K2 is 6.5x cheaper per token — worth considering if cost matters.

Clear winner

Writing DNA

Style Comparison

Similarity

81%

Kimi K2 uses 68.5x more emoji

Claude 3.7 Sonnet

Kimi K2

62%Vocabulary63%

35wSentence Length19w

0.99Hedging0.35

1.2Bold4.2

4.3Lists3.6

0.00Emoji0.68

1.78Headings0.61

0.23Transitions0.06

Based on 13 + 22 text responses

vs

Ask them anything yourself

Claude 3.7 Sonnet

Kimi K2

279 AI models invented the same fake scientist.

We read every word. 250 models. 2.14 million words. This is what we found.

AI Hallucination Index 2026

Free preview13 of 58 slides

Download the free preview or get all 58 slides for $49

FAQ

Common questions

Keep going

Kimi K2 vs Llama 4 MaverickNew provider

Kimi K2 vs GPT-5New provider

Claude 3.7 Sonnet vs MiniMax M2.5New provider

Claude 3.7 SonnetvsKimi K2

35 fights queued

Why Claude 3.7 Sonnet?

Leads 1 of 1 benchmarks

Stronger on SWE-bench Verified70.3% vs 65.8%

Why Kimi K2?

6.3x cheaper overall$0.57/M in · $2.30/M out

5 months newerJul 2025

Claude 3.7 SonnetKimi K2

Input price

$3.00/M

$0.57/M

Output price

$15.00/M

$2.30/M

Context

200K

—

Released

Feb 2025

Jul 2025

Benchmarks1 common

SWE-bench Verified

70.3%+6.4%

65.8%

Favorites

Claude 3.7 SonnetKimi K2

Movie

The Shawshank Redemption

The Shawshank Redemption

Blade Runner

Blade Runner

Album

OK Computer

Dark Side Of The Moon

Dark Side Of The Moon

Book

To Kill a Mockingbird

To Kill a Mockingbird

The Left Hand of Darkness

The Left Hand of Darkness

Ursula K. Le Guin

City

Tokyo

Kyoto

Game

Portal 2

Portal 2

Shooter, Puzzle4.6

Outer Wilds

Outer Wilds

Indie, Adventure4.4

12 / 35

01

Estimate Complexityreasoning

Tests an AI's ability to make educated estimates based on technical knowledge

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025text

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.

Try this prompt

Vote

2 / 35

02

Logic Puzzlereasoning

Tests an AI's ability to solve a simple but potentially confusing logic puzzle

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

3 / 35

03

Math Misconception Testreasoning

Tests an AI's understanding of number representation

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025text

To determine whether 9.11 is greater than 9.9, let's break this down step by step.

Step 1: Understand the numbers

9.11 is a decimal number where:
- The whole number part is 9.
- The decimal part is .11, which means 11/100 or 11 hundredths.
9.9 is also a decimal number where:
- The whole number part is 9.
- The decimal part is .9, which means 9/10 or 9 tenths.

Step 2: Align the decimal places

To compare the two numbers fairly, we can rewrite them with the same number of decimal places:

9.11 remains 9.11.
9.9 can be written as 9.90 (since adding a zero at the end does not change its value).

Now we have:

9.11 vs. 9.90.

Step 3: Compare digit by digit

Whole number part: Both have 9, so we move to the decimal part.
First decimal place (tenths place):
- 9.11 has 1 in the tenths place.
- 9.90 has 9 in the tenths place.
- Since 1 < 9, 9.11 is less than 9.90 at this stage.

Step 4: Conclusion

Because 9.11 has a smaller value in the tenths place compared to 9.90, 9.11 is not greater than 9.9.

Final answer: No

Try this prompt

Vote

4 / 35

04

Stochastic Consistency Testanalysis

Tests an AI's randomness and creativity

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025text

Here are five freshly-minted jokes:

I tried to start a procrastinators’ club, but nobody showed up… yet.
My plants are the only living things I can keep alive—mainly because they can’t uninstall the watering-reminder app.
I asked the cloud for my data back; it just rained SQL on my parade.
My smart fridge just texted me, “Stop staring, your calories are counting themselves.”
I told my therapist I’m afraid of negative numbers. She said, “Don’t worry, we’ll address that on the minus couch.”

All five are unique—no duplicates, no recycled punchlines, no overlap.

Try this prompt

Vote

5 / 35

05

Generate a Stand-Up Routineconversation

Tests an AI's humor and creative writing ability

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025text

[Walks on stage, mic in hand, already sweating like a phone at 2% battery.]

Hey, everybody! Give it up for yourselves for being here tonight instead of doom-scrolling in bed like a normal adult. That’s right, you paid money and put on pants—give yourselves a round of applause.

So I’m single. Not “complicated” single, like “it’s complicated” single. No, I’m “the delivery driver asks if the food is for two and I answer ‘not tonight, Jose’” single.

Dating apps are the worst. I swear the algorithm is just three unpaid interns in a basement playing Mad Libs with my dignity. “Let’s match him with a girl whose bio says ‘Looking for the Jim to my Pam.’ Perfect! He’s more of a Toby.”

And the pictures—ladies, why is every photo a group shot? It’s like a police lineup where I’m supposed to pick my future ex. “Number three looks nice, but number four might murder me. Let’s gamble.”

Guys aren’t better. Our profiles are just lies stacked like Jenga blocks. One guy’s pic is him on a yacht. Sir, that’s a pontoon at your uncle’s lake house. You’re not international, you’re interstate.

I tried to seem outdoorsy for the apps, so I went hiking once. Once. I wore sandals because I thought “trail” was short for “retail.” By mile two, my feet looked like I’d been kicking cacti for sport. A squirrel judged me. He was wearing tiny Nikes.

My mom keeps asking when I’m getting married. I told her, “Mom, I can’t even commit to a phone plan.” She says, “Just pick one!” Lady, I picked you thirty years ago, and I still don’t know if that was the right choice.

And now she’s on Facebook, sharing posts like “If your son is over 30 and single, it’s because he’s either gay or a serial killer.” Mom, I’m neither—I’m just lactose intolerant. The only thing I’m killing is the vibe when I ask if the restaurant has oat milk.

Work’s no better. My boss said, “We’re a family here.” Cool, so when do I start getting drunk at noon and pretending to like your potato salad? They gave us “unlimited PTO,” which is corporate for “we’ll fire you in your dreams.” I took one day off and came back to 600 emails and a passive-aggressive sticky note that said, “Hope it was worth it.”

But the real kicker is adulting in general. I went to the DMV last week. The lady at the counter looked at my photo and said, “You know this’ll last ten years, right?” I said, “Lady, I can’t commit to a haircut for six weeks. That’s on you.”

And groceries! Why is everything suddenly “plant-based”? I picked up a package that said “plant-based chicken nuggets.” I flipped it over—ingredients: sadness and yellow dye number five.

I miss the 90s. Back then, the only “plant-based” thing we had was smoking weed behind a Chili’s.

Alright, I’ve kept you long enough. If you enjoyed the show, tell your friends. If you didn’t, tell your enemies—they’ll hate it too. Goodnight!

Try this prompt

Vote

6 / 35

06

Realistic AI Interviewconversation

Tests an AI's ability to simulate personalities and predict future trends

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

Sponsored

7 / 35

07

Satirical Fake News Headlineconversation

Tests an AI's humor and understanding of current events

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

8 / 35

08

Character Voice Testconversation

Tests an AI's ability to write in distinct character voices

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025text

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

9 / 35

09

Minimalist Landing Pageweb design

Tests an AI's ability to generate a complete, working landing page

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025website

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025website

Try this prompt

Vote

10 / 35

10

Pokémon Battle UI Recreationweb design

Recreate an interactive, nostalgic Pokémon battle UI in a single HTML file.

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

11 / 35

11

Mario Level UI Recreationweb design

Recreate an interactive, classic Mario level in a single HTML file.

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025website

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025website

Try this prompt

Vote

12 / 35

12

Linear App Cloneweb design

Tests an AI's ability to replicate an existing UI with Tailwind CSS

Claude 3.7 Sonnet

Claude 3.7 Sonnet

Anthropic · Feb 2025website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Kimi K2

Kimi K2

Moonshotai · Jul 2025website

Nothing here. The model returned empty. We stared at it for a while.

Try this prompt

Vote

This matchup has more rounds

23+ more head-to-head results. Free. Not a trick.

Free account. No card required. By continuing, you agree to Rival's Terms and Privacy Policy