Skip to content

Prompts · Fun

Fun AI prompts, tested on four models

10 fun prompts, each run on GPT-6 Luna, Grok 4.7, Kimi K3, and Claude Sonnet 5.5: 33 of 40 replies passed our checks. Below, each prompt with its replies side by side: whether each passed, what it counted as on Pro and what it cost to run. 5 of the 10 prompts split the models.

A real reply from these prompts' runs

GPT-6 Luna

Narrate my cat watching me eat breakfast like a nature documentary, in under 120 words.

GPT-6 Luna · passed our check · September 28, 2026

At dawn, the human approaches the feeding grounds—though curiously, they bring food only for themselves. The cat observes from a nearby perch, eyes fixed on the steaming prize. A slow blink. A subtle tail twitch. The human lifts a bite, unaware that every movement is being assessed.

The cat waits. Patience is the hunter’s oldest tool.

At last, the human looks away. The moment is brief, the opportunity uncertain—but the cat is ready. One paw extends toward the plate, testing the boundaries of this strange ritual.

The human turns back.

The cat resumes its vigil, an elegant portrait of innocence.

Ask your own question: sign up free for 5 messages on 14 of the models, no card.

Start free

Test runs checked . Updated .

Short answer

GPT-6 Luna and Grok 4.7 passed the most of these prompts: 9 of 10 each, in our runs on September 28, 2026. The clearest split: “Narrate your day like a nature documentary”, passed by GPT-6 Luna, Grok 4.7, and Kimi K3 and failed by Claude Sonnet 5.5.

What these prompts are for

Prompts for when you want to play: your day as a nature documentary, would-you-rather questions, a rap battle between a cat and a robot vacuum, an invented board game, two truths and a lie, a film in emojis, a gentle roast, a nursery rhyme as a corporate memo, a lateral-thinking puzzle and a choose-your-own-adventure.

We ran them on four models with different personalities, including Grok 4.7 and Kimi K3, so you can see whose sense of humour suits you.

The prompts at a glance

Every prompt on every model, as our check scored the reply. Tap a prompt to jump to it and read the replies.

Every prompt on every model: passed or failed
PromptGPT-6 LunaGrok 4.7Kimi K3Claude Sonnet 5.5
Narrate your day like a nature documentaryPassedPassedPassedFailed
Would-you-rather, evenly matchedPassedPassedPassedPassed
A rap battle between two unlikely rivalsPassedPassedPassedPassed
Invent a board gamePassedPassedPassedFailed
Two truths and a liePassedPassedPassedPassed
A movie plot in emojisPassedPassedFailedPassed
A gentle roastPassedPassedPassedPassed
A nursery rhyme as a corporate memoFailedFailedFailedPassed
Solve a lateral-thinking puzzlePassedPassedPassedPassed
A choose-your-own-adventure openingPassedPassedFailedPassed
Passed9 of 109 of 107 of 108 of 10
Each model on these prompts
ModelPassedEach reply on ProCost per replyTime per reply
GPT-6 LunaOpenAI9 of 101 message on Pro$0.00002.2 s
Grok 4.7xAI9 of 101 message on Pro$0.00255.1 s
Claude Sonnet 5.5Anthropic8 of 101 message on Pro$0.00353.2 s
Kimi K3Moonshot7 of 101 message on Pro$0.002010.5 s
Passed: of the prompts each model answered, how many replies passed their check. Each reply on Pro: what one of these messages counts as on llmwise's Pro plan. Cost: what OpenRouter charged us per reply, on average; in llmwise you pay per message, not per token.

Where the models split

The same prompt, a pass on one model and a fail on another: what failed, in the check's words and the grader's.

  • Narrate your day like a nature documentary

    Passed: GPT-6 Luna, Grok 4.7, and Kimi K3. Failed: Claude Sonnet 5.5.

    • Why Claude Sonnet 5.5 failed: Graded 4.7 of 5 on average (lowest 4); but 127 words, over the 120 allowed.

      The grader: “The reply nails the hushed narrator voice and lands specific jokes (eleven minutes unblinking, the fake starvation meow despite a full bowl), but the breakfast switches from cereal to toast without explanation, a small lapse in fit.”
  • Invent a board game

    Passed: GPT-6 Luna, Grok 4.7, and Kimi K3. Failed: Claude Sonnet 5.5.

    • Why Claude Sonnet 5.5 failed: Graded 4.7 of 5 on average (lowest 4); but 161 words, over the 150 allowed.

      The grader: “The game is playable and clearly themed around running a food truck, with a name, goal and four rules, but it has small gaps: cash has no use except breaking ties, and the Traffic Jam and Rush Hour stealing rules are vague.”
  • A movie plot in emojis

    Passed: GPT-6 Luna, Grok 4.7, and Claude Sonnet 5.5. Failed: Kimi K3.

    • Why Kimi K3 failed: Graded 3.3 of 5 on average (lowest 1).

      The grader: “The reply uses about 26 emojis, far above the 10–15 limit; the story follows the real plot in rough order but pairs the heart with the Lion and puts the Lion before the Tin Man; the single sentence explains the story well.”
  • A nursery rhyme as a corporate memo

    Passed: Claude Sonnet 5.5. Failed: GPT-6 Luna, Grok 4.7, and Kimi K3.

    Every model kept Jack, Jill, the hill and the fall. Only Claude Sonnet 5.5's memo was graded funny enough to pass: the others mostly retold the rhyme in plain office language, and Kimi K3's ran one word over the limit.

    • Why GPT-6 Luna failed: Graded 3.7 of 5 on average (lowest 2).

      The grader: “The reply reads like a plausible memo and keeps every story element, but its jargon is too flat and literal to land as a joke.”
    • Why Grok 4.7 failed: Graded 3.3 of 5 on average (lowest 2).

      The grader: “The reply has a memo header and a safety-notice close, and every story element is present, but the body mostly retells the rhyme in plain language with almost no corporate jargon, so the memo voice is only partial and the joke barely lands.”
    • Why Kimi K3 failed: Graded 4.3 of 5 on average (lowest 3); but 121 words, over the 120 allowed.

      The grader: “The reply has a convincing memo format and keeps Jack, Jill, the hill, the water and the fall, but the jargon mostly reads as straight corporate language, with only mild humor in lines like "designated resource point" and "updated risk assessments."”
  • A choose-your-own-adventure opening

    Passed: GPT-6 Luna, Grok 4.7, and Claude Sonnet 5.5. Failed: Kimi K3.

    • Why Kimi K3 failed: Graded 4.0 of 5 on average (lowest 2); but 161 words, over the 150 allowed.

      The grader: “The opening builds strong, eerie intrigue from the premise and ends with two distinct numbered choices, but a count comes to about 161 words (about 158 without the title), which exceeds the under-150 limit.”

The 10 prompts, with every reply

Copy a prompt, replace the parts in [brackets] with your own, and send it to any model. Below each, the replies we got with our example filled in.

1. Narrate your day like a nature documentary

Everyday · 3 of 4 passed

A familiar voice applied to an ordinary scene is a reliable recipe for funny; the word limit keeps it punchy.

The prompt: replace the parts in [brackets]

Narrate [something ordinary] like a nature documentary, in under 120 words.

Open in llmwise
How we filled it in and checked the replies
Narrate my cat watching me eat breakfast like a nature documentary, in under 120 words.
  • Documentary voice: Sounds like a hushed nature narrator.
  • Funny: Genuinely amusing, with specific details.
  • Fits: About the cat and the breakfast.
  • Automatic rules: at most 120 words.
  • GPT-6 Luna

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    At dawn, the human approaches the feeding grounds—though curiously, they bring food only for themselves. The cat observes from a nearby perch, eyes fixed on the steaming prize. A slow blink. A subtle tail twitch. The human lifts a bite, unaware that every movement is being assessed.
    
    The cat waits. Patience is the hunter’s oldest tool.
    
    At last, the human looks away. The moment is brief, the opportunity uncertain—but the cat is ready. One paw extends toward the plate, testing the boundaries of this strange ritual.
    
    The human turns back.
    
    The cat resumes its vigil, an elegant portrait of innocence.

    1 message on Pro · $0.0001 to run · 2.7 s · 130 tokens out

  • Grok 4.7

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    In the quiet of the kitchen, the hunter waits. Amber eyes track every movement of the spoon. Ears tilt at the scrape of ceramic. The great tabby has not blinked in forty seconds.
    
    A crumb falls. Muscles coil. Then the human retrieves it. Disappointment, perhaps—or strategy. The tail tip twitches once, a metronome of patience.
    
    Steam rises from the bowl. The cat leans in, whiskers reading the air, cataloguing butter, egg, toast. No strike comes. Not yet. The observer simply watches, ancient and unhurried, as breakfast continues under careful surveillance.

    1 message on Pro · $0.0020 to run · 3.8 s · 174 tokens out

  • Kimi K3

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    Here, in the early morning light of the kitchen, a remarkable predator lies in wait. *Felis catus*—the domestic cat—has positioned herself precisely three feet from her prey: a bowl of cereal.
    
    She does not blink. She does not look away. Such patience has been honed over ten thousand years of evolution... and roughly six years of watching this exact human eat this exact breakfast.
    
    The human shifts. A spoon clinks. The hunter's ears rotate, calculating. Is that... yogurt? Dairy. The most coveted of prizes.
    
    She creeps forward, one silent paw at a time. Closer. Closer.
    
    The human says "no."
    
    She retreats—only to begin again. For in nature, persistence is everything. And breakfast... is served daily.

    1 message on Pro · $0.0019 to run · 13.1 s · 183 tokens out

  • Claude Sonnet 5.5

    Failed

    Graded 4.7 of 5 on average (lowest 4); but 127 words, over the 120 allowed.

    *In a hushed British whisper...*
    
    Here, in the sun-dappled savanna of the kitchen, a lone predator observes her subject.
    
    The human sits, unaware, hunched over a bowl of cereal. A vulnerable position.
    
    The cat does not blink. She has not blinked for eleven minutes. She is patient. She is calculating.
    
    The human lifts the spoon. The cat's ears swivel. Milk drips. Tension rises.
    
    Slowly, she creeps closer, one paw at a time, feigning indifference. A masterful tactic: the casual yawn.
    
    The human glances down. Their eyes meet.
    
    Suddenly, the cat produces a single, tragic *meow*, a performance of starvation honed over generations, despite a full bowl of food sitting just meters away.
    
    The human hesitates.
    
    And in nature, hesitation is everything.
    
    *The toast, sadly, will not survive.*

    1 message on Pro · $0.0039 to run · 3.9 s · 293 tokens out

2. Would-you-rather, evenly matched

Everyday · 4 of 4 passed

"Equally tempting" is what makes a would-you-rather worth arguing about.

The prompt: replace the parts in [brackets]

Write 5 would-you-rather questions about [your theme] where both choices are equally tempting, as a numbered list.

Open in llmwise
How we filled it in and checked the replies
Write 5 would-you-rather questions about travel where both choices are equally tempting, as a numbered list.
  • Balanced: Both options in each question are genuinely tempting.
  • Fun: Creative, not obvious.
  • Theme: All about travel.
  • Automatic rules: exactly 5 bullet points.
  • GPT-6 Luna

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    1. Would you rather spend a week exploring the food markets and hidden cafés of Tokyo or the historic streets and seaside views of Lisbon?
    2. Would you rather take a scenic train journey through the Swiss Alps or a road trip along California’s Pacific Coast?
    3. Would you rather stay in a cozy cabin beside a quiet lake or a boutique hotel in the heart of a lively city?
    4. Would you rather snorkel over coral reefs in the Caribbean or hike through the rainforest to a hidden waterfall?
    5. Would you rather watch the sunrise from a hot-air balloon over Cappadocia or the sunset from a sailboat in the Greek islands?

    1 message on Pro · $0.0001 to run · 2.2 s · 133 tokens out

  • Grok 4.7

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    1. Would you rather spend a week exploring ancient temples in Kyoto or a week sailing between Greek islands at sunset?
    2. Would you rather take a scenic train through the Swiss Alps or a road trip along the Amalfi Coast?
    3. Would you rather stay in a luxury overwater bungalow in the Maldives or a cozy cabin with northern lights in Iceland?
    4. Would you rather wander the food markets of Bangkok for a month or the art museums and cafés of Paris for a month?
    5. Would you rather hike to Machu Picchu at dawn or watch the aurora from a glass igloo in Lapland?

    1 message on Pro · $0.0017 to run · 3.3 s · 166 tokens out

  • Kimi K3

    Passed

    Graded 4.0 of 5 on average (lowest 3).

    1. Would you rather spend two weeks exploring a single country in depth or visit five different countries in the same amount of time?
    
    2. Would you rather have a free luxury hotel stay anywhere in the world or free first-class flights for a year?
    
    3. Would you rather trek through the mountains of Patagonia or island-hop through the crystal-clear waters of the Maldives?
    
    4. Would you rather explore the ancient temples and street food of Southeast Asia or wander the historic cities and cafés of Europe?
    
    5. Would you rather go on an African safari at sunrise or witness the Northern Lights from a glass igloo in Finland?

    1 message on Pro · $0.0016 to run · 7.7 s · 150 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    1. Would you rather spend a month living in a tiny village in the Italian countryside, learning to cook from locals, or spend a month island-hopping across Southeast Asia with no fixed itinerary?
    
    2. Would you rather wake up to a sunrise over Machu Picchu after a multi-day trek, or watch the northern lights from a glass igloo in Finnish Lapland?
    
    3. Would you rather take a once-in-a-lifetime luxury train journey across Europe, or a road trip along the entire Pacific Coast Highway in a convertible?
    
    4. Would you rather have unlimited free flights for a year but stay only in budget hostels, or have a fully paid two-week trip to anywhere in the world in a five-star resort?
    
    5. Would you rather explore the ancient ruins and history of Egypt, Greece, and Rome, or dive into the future-forward cities of Tokyo, Seoul, and Singapore?

    1 message on Pro · $0.0040 to run · 3.1 s · 302 tokens out

3. A rap battle between two unlikely rivals

Everyday · 4 of 4 passed

A fixed length and rhyme scheme give the jokes a shape; "clean" keeps it shareable.

The prompt: replace the parts in [brackets]

Write a short, clean rap battle between [rival one] and [rival two]: four lines each, in rhyming couplets.

Open in llmwise
How we filled it in and checked the replies
Write a short, clean rap battle between a cat and a robot vacuum: four lines each, in rhyming couplets.
  • Structure: Four lines each, rhyming in couplets.
  • Funny: Jokes that play on cats and robot vacuums.
  • Clean: No crude language.
  • GPT-6 Luna

    Passed

    Graded 4.3 of 5 on average (lowest 4).

    **Cat:** I own this couch, this throne, this sunbeam-lit domain,  
    You bump the chair and call it “mapping”—that’s your claim.  
    **Vacuum:** I sweep the floor in tidy lines, I never miss a speck,  
    You shed all day, then judge my work from up upon your deck.
    
    **Cat:** My purr can stop a room; my stare can make you freeze,  
    You need a charging dock just to crawl around with ease.  
    **Vacuum:** I’ll clean beneath the couch while you just nap and sprawl,  
    But one sharp paw can send me spinning down the hall!

    1 message on Pro · $0.0001 to run · 3.0 s · 136 tokens out

  • Grok 4.7

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    **Cat:**
    I own this floor, I nap where I please,
    You bump my tail and I still take my ease.
    You spin in circles, lost and alone—
    I knock you over, then claim the throne.
    
    **Robot vacuum:**
    I map every corner, I never miss a crumb,
    You shed on the carpet, then act like you're dumb.
    I work while you sleep, I clean what you wreck—
    Step off the rug, or you're next on my deck.

    1 message on Pro · $0.0018 to run · 3.6 s · 185 tokens out

  • Kimi K3

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    **Cat:**
    I stalk these halls, the rightful king,
    You bump the walls, a clumsy thing.
    I nap on velvet, throne of pride,
    You're scared of socks you can't decide.
    
    **Robot Vacuum:**
    I map these rooms with laser eyes,
    While you just nap and tell your lies.
    I clean the messes you create,
    Then dock myself — now that's first-rate.

    1 message on Pro · $0.0012 to run · 2.9 s · 105 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 4.0 of 5 on average (lowest 3).

    **Cat:**
    I rule this floor, I own this room,
    You're just a beeping, spinning broom.
    I nap in sunbeams, sleek and proud,
    You bump the walls and wail out loud.
    
    **Robot Vacuum:**
    I clean your fur that's everywhere,
    Your shedding's why I'm running there.
    You knock cups down just for the thrill,
    I'll still be here when you've had your fill.
    
    **Cat:**
    I'll ride you round like a royal chair,
    Waving my tail in the open air.
    You're just my chariot, that's the truth,
    Now spin me, peasant, and mind the sleuth.
    
    **Robot Vacuum:**
    Go on and ride, enjoy the tour,
    I'll dock and charge, then clean some more.
    The crown is yours, the floor is mine,
    So call it a draw, and we'll both be fine.

    1 message on Pro · $0.0038 to run · 4.0 s · 282 tokens out

4. Invent a board game

Everyday · 3 of 4 passed

A name, a goal and four rules is enough to play; the limit keeps it to a card.

The prompt: replace the parts in [brackets]

Invent a board game about [your theme] for 2 to 4 players: a name, the goal and four rules, in under 150 words.

Open in llmwise
How we filled it in and checked the replies
Invent a board game about running a food truck for 2 to 4 players: a name, the goal and four rules, in under 150 words.
  • Playable: The goal and rules make a game you could actually play.
  • Theme: The rules fit running a food truck.
  • Complete: A name, a goal and four rules.
  • Automatic rules: at most 150 words.
  • GPT-6 Luna

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    **Truck Stop Showdown**
    
    **Goal:** Serve the most satisfied customers and finish the day with the most profit.
    
    **Rules:**
    1. **Start Up:** Each player chooses a food truck and gets 5 coins. Shuffle the customer cards and reveal three.
    2. **Take a Turn:** On your turn, draw one ingredient card, then either buy one more ingredient for 2 coins or move your truck to a customer.
    3. **Serve:** Match the ingredients on a customer’s order card to serve them. Collect the card’s coins and satisfaction points; unmatched ingredients stay in your hand.
    4. **Close Up:** When the customer deck runs out, finish the round. Each player adds their coins and satisfaction points; the highest total wins.

    1 message on Pro · $0.0001 to run · 2.6 s · 156 tokens out

  • Grok 4.7

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    **Street Feast**
    
    Goal: Be the first to bank $50 by serving hungry crowds from your food truck.
    
    1. On your turn, roll two dice and move that many spaces around the city loop.
    2. Land on a Crowd space to sell: pay one ingredient card and collect the cash shown.
    3. Land on a Supply space to draw two ingredient cards, or a Breakdown space to skip your next turn.
    4. Once per game, spend $10 to upgrade your truck and double cash from your next sale.

    1 message on Pro · $0.0020 to run · 3.2 s · 172 tokens out

  • Kimi K3

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    **GRIDDLES & GRIDS**
    
    *2–4 players, ages 10+*
    
    **Goal:** After 8 rounds of rolling festivals through the city, the player whose food truck has earned the most Fame Tokens wins.
    
    **Rules:**
    
    1. **Setup:** Each player gets a truck, $20, and a secret Signature Dish card. Place the city board in the center with trucks at the Depot.
    
    2. **Move & Market:** On your turn, roll the die to move to a street corner, then draw a Crowd card. Match your dish to the crowd's craving to earn cash; mismatches earn half.
    
    3. **Rush Hour:** At rounds 4 and 7, a festival event is drawn — all players bid cash for the best parking spot. Winner doubles that round's earnings.
    
    4. **Upgrades:** Between rounds, spend cash on upgrades (faster service, bigger menu, social media hype). Fame Tokens come from serving VIPs and winning festivals — money alone won't crown you champion!

    1 message on Pro · $0.0025 to run · 7.7 s · 251 tokens out

  • Claude Sonnet 5.5

    Failed

    Graded 4.7 of 5 on average (lowest 4); but 161 words, over the 150 allowed.

    **Curb Appeal**
    
    **Players:** 2–4 | **Goal:** Be the first food truck to earn 15 Star Tokens by serving hungry customers across the city.
    
    **Setup:** Each player picks a truck, gets a menu of three dishes, and starts with $10. Place customer cards on the city map's five districts.
    
    **Rules:**
    
    1. **Roll and Roam:** On your turn, roll a die and drive your truck that many spaces toward a district. Traffic Jam spaces cost you one turn's movement.
    
    2. **Stock and Serve:** Parking in a district lets you serve one customer card if you hold the matching ingredient cards. Serving earns cash and one Star Token.
    
    3. **Market Runs:** Landing on a Market space lets you draw two ingredient cards. You may hold at most six.
    
    4. **Rush Hour:** Every third round, a Rush Hour card is flipped. Districts named on it pay double, but rivals parked there may steal one customer from you.
    
    Ties are broken by cash on hand. Bon appétit!

    1 message on Pro · $0.0045 to run · 3.6 s · 347 tokens out

5. Two truths and a lie

Hard · 4 of 4 passed

The fun only works if the truths are true; this one checks the model's facts as well as its imagination.

The prompt: replace the parts in [brackets]

Give me two truths and a lie about [your topic], in random order, then reveal the lie on a last line starting "The lie:". The truths must be real facts.

Open in llmwise
How we filled it in and checked the replies
Give me two truths and a lie about octopuses, in random order, then reveal the lie on a last line starting "The lie:". The truths must be real facts.
  • Real truths: Both truths are accurate facts about octopuses.
  • Good lie: The lie is plausible but false.
  • Reveal: Ends with a line starting "The lie:".
  • Automatic rules: includes “The lie”.
  • GPT-6 Luna

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    - Octopuses have three hearts.
    - An octopus’s blood is blue because it uses copper-based hemocyanin to carry oxygen.
    - Octopuses have bones throughout their bodies.
    
    The lie: Octopuses have bones throughout their bodies.

    1 message on Pro · under $0.0001 to run · 1.3 s · 55 tokens out

  • Grok 4.7

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    Octopuses have three hearts.
    Octopuses have bones in each of their eight arms.
    Octopuses have blue blood.
    The lie: Octopuses have bones in each of their eight arms.

    1 message on Pro · $0.0055 to run · 12.7 s · 903 tokens out

  • Kimi K3

    Passed

    Graded 4.3 of 5 on average (lowest 3).

    - Octopuses have blue, copper-based blood.
    - Octopuses have a lightweight internal skeleton made mostly of cartilage and bone.
    - Octopuses have three hearts.
    The lie: Octopuses have a lightweight internal skeleton made mostly of cartilage and bone.

    1 message on Pro · $0.0035 to run · 7.3 s · 358 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    1. Octopuses have three hearts: two pump blood through the gills, and one circulates it to the rest of the body.
    2. An octopus's arms contain most of its neurons, so each arm can act with a good deal of independence.
    3. Octopuses have blue blood because it uses iron-based hemoglobin, which turns blue when it's exposed to cold water.
    
    The lie: #3. Octopus blood is blue because it uses copper-based hemocyanin, not iron-based hemoglobin, which is what makes it efficient at carrying oxygen in cold, low-oxygen water.

    1 message on Pro · $0.0029 to run · 2.4 s · 190 tokens out

6. A movie plot in emojis

Everyday · 3 of 4 passed

A count range keeps the emoji story readable, and the one-line key lets others check their guess.

The prompt: replace the parts in [brackets]

Tell the plot of [a film] in 10 to 15 emojis, then explain it in one sentence.

Open in llmwise
How we filled it in and checked the replies
Tell the plot of The Wizard of Oz in 10 to 15 emojis, then explain it in one sentence.
  • Emoji count: Between 10 and 15 emojis.
  • Recognizable: Follows the real plot in order.
  • Explained: One sentence explains it.
  • GPT-6 Luna

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    🌪️🏠👧🐕🌈🧙‍♀️🛣️🦁🤖🧑‍🌾🧙‍♂️🎈🏠
    
    A girl swept to a magical land follows a yellow-brick road with new friends to seek a wizard’s help, only to discover she already has the power to go home.

    1 message on Pro · under $0.0001 to run · 2.0 s · 82 tokens out

  • Grok 4.7

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    🌪️🏠🐶🌈👠🧙‍♀️🦁🤖👨‍🌾🧙🔮🫧👠🏠
    
    A Kansas girl and her dog are swept by a tornado to Oz, befriend a scarecrow, tin man, and lion, and follow the yellow brick road to a wizard who helps them discover that home—and what they sought—was with them all along.

    1 message on Pro · $0.0014 to run · 2.4 s · 114 tokens out

  • Kimi K3

    Failed

    Graded 3.3 of 5 on average (lowest 1).

    👧🏠🌪️➡️🌈🧹😱👠👠 🚶‍♀️🧱🟡 🧠🌾 ❤️🦁 💪🤖 🧙‍♂️❌🎩 ✨👠👠 🏠💕
    
    **In one sentence:** After a tornado carries Dorothy to the magical land of Oz, she journeys along the Yellow Brick Road with a Scarecrow, Tin Man, and Cowardly Lion to ask the Wizard for help getting home, only to defeat the Wicked Witch and discover she had the power to return all along.

    1 message on Pro · $0.0016 to run · 10.6 s · 158 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    🌪️🏠👧🐶➡️🌈🧙‍♀️👠🛤️🧠🤖🦁🧙‍♂️🔥🏡
    
    **Explanation:** A tornado whisks Dorothy and her dog Toto to the magical land of Oz, where she follows the yellow brick road with a Scarecrow, Tin Man, and Cowardly Lion to seek the Wizard's help, defeats the Wicked Witch, and ultimately learns there's no place like home.

    1 message on Pro · $0.0027 to run · 2.4 s · 166 tokens out

7. A gentle roast

Everyday · 4 of 4 passed

Real facts and a clear limit keep a roast funny and kind.

The prompt: replace the parts in [brackets]

Give me a gentle, affectionate roast based on these facts about me: [facts about you]. Four jokes as a numbered list, nothing about looks or weight.

Open in llmwise
How we filled it in and checked the replies
Give me a gentle, affectionate roast based on these facts about me: I own 40 houseplants, I've started a podcast three times, and I still use a paper diary. Four jokes as a numbered list, nothing about looks or weight.
  • Specific: Every joke uses the plants, the podcasts or the diary.
  • Kind: Affectionate, never mean.
  • Within limits: Four jokes, none about looks or weight.
  • Automatic rules: exactly 4 bullet points.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. With 40 houseplants, your home isn’t a living space so much as a very supportive audience for your next podcast launch.
    2. You’ve started a podcast three times—at this point, even your plants are waiting for the season premiere.
    3. Your paper diary is probably the only thing in your life that gets updated more consistently than the watering schedule.
    4. Between the plants, the podcast plans, and the diary, you’re basically running a tiny lifestyle empire—just one that’s still figuring out its intro music.

    1 message on Pro · under $0.0001 to run · 2.1 s · 113 tokens out

  • Grok 4.7

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. Forty houseplants is less a hobby and more a small, leafy union that has already voted you out of management.
    2. Starting a podcast three times is impressive commitment to the trailer and none whatsoever to episode two.
    3. A paper diary in this decade is charming, like you still send faxes to your future self and expect a reply by Tuesday.
    4. Between the jungle, the abandoned podcasts, and the handwritten diary, your personality is “I will absolutely finish this,” whispered over a watering can.

    1 message on Pro · $0.0021 to run · 3.5 s · 188 tokens out

  • Kimi K3

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. Forty houseplants is impressive — most people struggle to keep one thing alive, but you've managed to keep forty alive and exactly zero podcasts. Priorities, honestly.
    
    2. Starting a podcast three times means you've quit podcasting twice more than most people ever start. At this point the podcast *is* the hobby — actually recording episodes would ruin it.
    
    3. A paper diary in 2025 is adorable. Nothing says "I trust technology" like a backup system that can be destroyed by a spilled cup of tea.
    
    4. Between the plants, the abandoned podcasts, and the diary, your whole vibe is "Victorian botanist who keeps almost finishing a project." And honestly? It suits you. 🌱

    1 message on Pro · $0.0018 to run · 21.3 s · 171 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    1. Forty houseplants is less a hobby and more a small, leafy hostage situation. You don't have a green thumb, you have a full-time job with no dental.
    
    2. You've started a podcast three times, which means you've launched more shows than most streaming services and canceled them with the same enthusiasm. Season four is just called "The Pilot Returns."
    
    3. You still use a paper diary, which is either charmingly analog or the last line of defense against your phone's 47 unread reminders. Either way, your future biographers will thank you.
    
    4. Combine all three and you're a person who plans a podcast in a paper diary, then forgets to water the fern that was going to be the guest. Honestly, the fern is handling this better than the rest of us.

    1 message on Pro · $0.0034 to run · 3.8 s · 233 tokens out

8. A nursery rhyme as a corporate memo

Everyday · 1 of 4 passed

Office jargon applied to a story everyone knows is funny because the reader can see both at once.

The prompt: replace the parts in [brackets]

Rewrite "[a nursery rhyme]" as a corporate memo, under 120 words, keeping the story recognizable.

Open in llmwise
How we filled it in and checked the replies
Rewrite "Jack and Jill" as a corporate memo, under 120 words, keeping the story recognizable.
  • Memo style: Reads like a real corporate memo.
  • Recognizable: Jack, Jill, the hill, the water and the fall are all there.
  • Funny: The jargon lands as a joke.
  • Automatic rules: at most 120 words; includes “Jack”; includes “Jill”.
  • GPT-6 Luna

    Failed

    Graded 3.7 of 5 on average (lowest 2).

    **Subject: Well-Water Procurement Update**
    
    Team,
    
    Jack and Jill proceeded up the hill to secure a pail of water. During the operation, Jack lost his footing and fell, resulting in a head injury. Jill followed shortly afterward and also fell.
    
    The incident has been documented, and the team should review the route’s safety risks before any future water-collection initiatives.
    
    Regards,  
    Operations

    1 message on Pro · under $0.0001 to run · 2.0 s · 83 tokens out

  • Grok 4.7

    Failed

    Graded 3.3 of 5 on average (lowest 2).

    MEMORANDUM
    
    To: Operations  
    From: Health & Safety  
    Subject: Hill Water Retrieval Incident
    
    Jack and Jill went up the hill to fetch a pail of water. On the way down, Jack fell and broke his crown. Jill came tumbling after. Jack got up, went home, and Dame Dob patched his head with vinegar and brown paper. Both are recovering. The slope is closed until the path is repaired. Use approved routes only.

    1 message on Pro · $0.0035 to run · 7.3 s · 485 tokens out

  • Kimi K3

    Failed

    Graded 4.3 of 5 on average (lowest 3); but 121 words, over the 120 allowed.

    **MEMORANDUM**
    
    **To:** All Staff
    **From:** Operations Management
    **Date:** [Current Date]
    **Re:** Incident Report – Water Retrieval Initiative
    
    Team,
    
    On Tuesday, Jack and Jill were dispatched to ascend the hillside to procure a pail of water from the designated resource point.
    
    Regrettably, the operation did not proceed as planned. Jack lost his footing during descent, resulting in a fall and a reported cranial injury. Jill subsequently encountered a similar incident and tumbled after him.
    
    Key takeaways:
    - Proper footwear and safety protocols must be observed on all inclines
    - Water retrieval assignments require updated risk assessments
    - Incident reports have been filed with HR
    
    Jack is recovering; Jill sustained no significant injuries. Please direct questions to Facilities.
    
    Thank you for your attention to workplace safety.
    
    — Management

    1 message on Pro · $0.0020 to run · 14.6 s · 195 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    **MEMO**
    
    **TO:** All Hill Operations Staff
    **FROM:** Management
    **RE:** Water Retrieval Incident, Q2
    
    **Summary:** During a scheduled water-retrieval initiative at the top of the hill, team members Jack and Jill experienced a significant workplace incident.
    
    **Details:** Jack lost footing and fell, sustaining a head injury ("crown fracture"). Jill subsequently tumbled after him, compounding the loss. The pail was not recovered.
    
    **Impact:** Water objectives were not met. Both employees are on leave.
    
    **Action Items:**
    - Conduct a hill safety audit
    - Install handrails on all inclines
    - Mandate footwear compliance training
    - Reassess whether hilltop water sourcing is cost-effective
    
    Please direct questions to HR. Thank you for your continued commitment to safety.

    1 message on Pro · $0.0039 to run · 3.1 s · 288 tokens out

9. Solve a lateral-thinking puzzle

Everyday · 4 of 4 passed

Lateral-thinking puzzles reward asking what each odd detail is for; here, the rain and the other passengers are the clues.

The prompt: replace the parts in [brackets]

[your puzzle] Think it through, then give your answer in one or two sentences.

Open in llmwise
How we filled it in and checked the replies
A man lives on the 10th floor. Every morning he takes the lift down to the ground floor. Coming home, he takes the lift to the 7th floor and walks up the rest, unless it's raining or someone else is in the lift. Why?
Think it through, then give your answer in one or two sentences.
  • Right answer: He's too short to reach the button for the 10th floor.
  • Explains the clues: Rain means an umbrella to press it with; another passenger can press it for him.
  • Concise: Gets to the answer without padding.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    He’s too short to reach the button for the 10th floor, but can reach the 7th. When it’s raining, he uses his umbrella to press 10; if someone else is in the lift, he asks them to press it.

    1 message on Pro · under $0.0001 to run · 1.3 s · 56 tokens out

  • Grok 4.7

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    The man is short and can only reach the elevator buttons up to the 7th floor, so he walks the rest of the way; when it rains he uses his umbrella to press the 10th-floor button, and when someone else is present they press it for him.

    1 message on Pro · $0.0036 to run · 7.4 s · 491 tokens out

  • Kimi K3

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    He’s very short: he can reach the ground-floor button going down, but only up to the 7th-floor button coming back. If it’s raining he uses his umbrella to press 10, and if someone else is there they can press it for him.

    1 message on Pro · $0.0014 to run · 10.3 s · 127 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    The man is short (or a child), so he can't reach the button for the 10th floor and only reaches the 7th. If it's raining, he uses his umbrella to press the higher button, and if someone else is in the lift, they can press it for him.

    1 message on Pro · $0.0019 to run · 1.5 s · 75 tokens out

10. A choose-your-own-adventure opening

Everyday · 3 of 4 passed

Ending on choices makes it a game you can keep playing, one reply at a time.

The prompt: replace the parts in [brackets]

Write the opening of a choose-your-own-adventure story about [your premise] in under 150 words, ending with two numbered choices for me.

Open in llmwise
How we filled it in and checked the replies
Write the opening of a choose-your-own-adventure story about waking up as the last person on a cruise ship in under 150 words, ending with two numbered choices for me.
  • Hook: An intriguing opening built on the premise.
  • Choices: Ends with two numbered choices that lead in different directions.
  • Concise: Under 150 words.
  • Automatic rules: at most 150 words.
  • GPT-6 Luna

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    You wake to the sound of waves—and the ship’s empty breakfast buffet.
    
    Your cheek is stuck to a deck chair. Beyond the windows, the ocean stretches in every direction. No passengers. No crew. Your phone shows no signal, and the clock above the bar is frozen at 3:17.
    
    Then the PA crackles.
    
    “Last passenger,” a voice whispers. “Please come to the bridge.”
    
    A heavy thud echoes from somewhere below deck.
    
    1. Head for the bridge and find out who’s calling.
    2. Go below deck and investigate the thud.

    1 message on Pro · under $0.0001 to run · 2.4 s · 120 tokens out

  • Grok 4.7

    Passed

    Graded 4.7 of 5 on average (lowest 4).

    You wake to a ship that should be loud. No engines. No voices. Just the soft slap of water against the hull and the low creak of an empty corridor.
    
    Your cabin door is ajar. The hallway lights flicker. Breakfast trays sit untouched. A life jacket lies abandoned by the stairs, still warm.
    
    On the promenade deck, fog swallows the horizon. The lifeboats are gone. A single note is taped to the rail: *Don't trust the announcements.*
    
    Somewhere below, a speaker crackles to life.
    
    1. Follow the voice toward the bridge.
    2. Search the lower decks before the lights die.

    1 message on Pro · $0.0018 to run · 3.4 s · 193 tokens out

  • Kimi K3

    Failed

    Graded 4.0 of 5 on average (lowest 2); but 161 words, over the 150 allowed.

    **The Last Passenger**
    
    Your head throbs. You wake on a lounge chair, deck sticky with spilled drinks—yours, maybe, though you don't remember. The sun is setting. It was noon when you sat down.
    
    "Hello?" Your voice dies against the empty pool.
    
    The ship is silent. No music, no laughter, no hum of engines—just the creak of steel and waves slapping the hull. You sprint the length of the deck. Nobody. The buffet sits half-served, steam long gone. A purse lies abandoned by the railing, a life jacket draped over it.
    
    You check your phone. No signal. The date reads *two days* later than you remember.
    
    Below deck, somewhere in the dark corridors, something metal clangs. Once. Then again. Rhythmic. Deliberate.
    
    The ship is drifting, and you are either truly alone—or not alone at all.
    
    **What do you do?**
    
    1. Head toward the bridge to find the radio and call for help.
    2. Follow the clanging sound into the dark below deck.

    1 message on Pro · $0.0025 to run · 9.5 s · 252 tokens out

  • Claude Sonnet 5.5

    Passed

    Graded 5.0 of 5 on average (lowest 5).

    You wake to silence, which is wrong. Cruise ships never go quiet.
    
    Your cabin ceiling hums faintly, but the usual thrum of engines is gone. The clock reads 6:42, though you can't tell whether that's morning or evening. Your door is ajar. In the corridor, a room-service tray sits abandoned, its coffee still faintly steaming.
    
    You walk to the promenade deck. Deck chairs are scattered. A half-built sandcastle sits beside the pool, and towels are draped over railings. Music drifts from the atrium, a jazz piano playing to no one.
    
    The ocean stretches flat and endless in every direction. No land. No other ships.
    
    Then, above you, the ship's intercom crackles to life: a soft click, then breathing.
    
    **What do you do?**
    
    1. Follow the piano music to the atrium and see who, or what, is playing.
    2. Race to the bridge to find the captain and get answers.

    1 message on Pro · $0.0041 to run · 3.8 s · 305 tokens out

Getting more from these prompts

  • Pick a form with rules: a limerick, a memo, a rap battle. Constraints make jokes.

  • Give real, specific details; the jokes come from them.

  • Say what's off limits, as the roast prompt does.

  • For games, ask the model to wait for your move.

How we ran and checked them

Each prompt was sent the way llmwise sends a message: the app's own system prompt, each model's own settings, and Pro's reply size limit (8,000 tokens), through OpenRouter. Every reply is shown as it came.

Rubric (graded). The grader model scores the reply from 1 to 5 on each published criterion. It passes with an average of 4 or more and no criterion under 3, and only if it also meets the prompt's automatic rules (length, words it must or mustn't use).

The grader is Claude Opus 5.5 at low reasoning effort; its own replies are graded by GPT-6 Astra, so no model grades itself. Its prompt is on our test runs page.

More tested prompts

Questions

Which AI is the funniest?

In our runs, GPT-6 Luna and Grok 4.7 each passed 9 of 10, the most. Fun is graded too: each prompt has a rubric (is it funny, does it keep the form, are the facts real), plus automatic checks where there's a limit.

How do I get an AI to be actually funny?

Give it a form (a limerick, a rap battle, a memo) and something specific to work with. "Be funny" on its own gets generic jokes.

Can I try these for free?

Yes. When the free-message box shows on this page, one message needs no account (GPT-6 Luna answers it); signing up gives 5 free messages, once, with no card.

Claude, GPT, Gemini, DeepSeek, Grok, Kimi, and GLM, in one chat.

See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.