Prompts · ChatGPT

ChatGPT prompts, tested on GPT-6

10 ChatGPT prompts, each run on GPT-6 Luna, GPT-6 Sol, GPT-6 Astra, and Claude Sonnet 5: 34 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. 4 of the 10 prompts split the models.

A real reply from these prompts' runs

GPT-6 Luna

Explain how a vaccine teaches your body to fight germs like I'm five. Under 100 words, no jargon, and use one everyday comparison.

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

A vaccine is like a practice alarm for your body’s guards. It shows them a safe clue about a germ, so they can learn what to look for. If the real germ shows up later, your guards can spot it faster and fight it off.

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

Start free

Test runs checked September 28, 2026. Updated September 29, 2026.

Short answer

GPT-6 Sol and GPT-6 Astra passed the most of these prompts: 9 of 10 each, in our runs on September 28, 2026. The clearest split: “Act as a career coach, one question at a time”, passed by Claude Sonnet 5 and failed by GPT-6 Luna, GPT-6 Sol, and GPT-6 Astra.

What these prompts are for

These are the everyday ChatGPT prompt patterns: act as someone, explain it simply, summarize, brainstorm, turn notes into a table, rewrite, weigh a decision, plan a day, get honest criticism and write in your own voice. Each is written the way that gets the best answer, with the details a model needs and a format it can't misread.

We ran each one on three GPT-6 models and on Claude Sonnet 5, so you can see what the same prompt gets from a bigger or a smaller model, and from another company's.

ChatGPT's pricing page lists OpenAI's GPT-6 Astra, Sol and Luna on its paid plans. We ran these prompts on those three models in llmwise, not in the ChatGPT app, so the replies show the models without ChatGPT's own memory, tools or settings.

Source: ChatGPT pricing, checked September 28, 2026.

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
Prompt	GPT-6 Luna	GPT-6 Sol	GPT-6 Astra	Claude Sonnet 5
Act as a career coach, one question at a time	Failed	Failed	Failed	Passed
Explain it like I'm five	Passed	Passed	Passed	Failed
Three bullets and a TL;DR	Passed	Passed	Passed	Passed
Ten ideas within a budget	Passed	Passed	Passed	Passed
Meeting notes into a table, no guessing	Passed	Passed	Passed	Passed
Firm but polite, every fact kept	Passed	Passed	Passed	Passed
Pros and cons with a verdict	Failed	Passed	Passed	Passed
Plan a day in time slots	Passed	Passed	Passed	Failed
A skeptical investor, no praise	Passed	Passed	Passed	Passed
Write it in my style	Passed	Passed	Passed	Passed
Passed	8 of 10	9 of 10	9 of 10	8 of 10

Each model on these prompts
Model	Passed	Each reply on Pro	Cost per reply	Time per reply
GPT-6 SolOpenAI	9 of 10	1 message on Pro	$0.0012	2.5 s
GPT-6 AstraOpenAI	9 of 10	1 message on Pro	$0.0066	5.6 s
GPT-6 LunaOpenAI	8 of 10	1 message on Pro	$0.0001	2.1 s
Claude Sonnet 5Anthropic	8 of 10	1 message on Pro	$0.0046	4.7 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.

Act as a career coach, one question at a time

Passed: Claude Sonnet 5. Failed: GPT-6 Luna, GPT-6 Sol, and GPT-6 Astra.

All three GPT-6 models asked two things joined by "and" (what you enjoy, and what you'd change), the usual way a model breaks "one at a time". Claude Sonnet 5 asked one question, about the move from the wards into health tech.

Why GPT-6 Luna failed: Graded 3.3 of 5 on average (lowest 2).

The grader: “The reply bundles two questions into one sentence ('what you enjoy most' and 'what you'd most like to change'), and it could be put to anyone in any job without mentioning ward nursing or health tech, though it correctly gives no advice.”

Why GPT-6 Sol failed: Graded 3.7 of 5 on average (lowest 3).

The grader: “The reply gives no advice and waits properly, but it joins two questions with 'and' (what they enjoy and what they'd like less of), and although it mentions nursing work it is a fairly generic career question that ignores health tech and the lack of coding skills.”

Why GPT-6 Astra failed: Graded 3.7 of 5 on average (lowest 3).

The grader: “The reply gives no advice and names the nursing-to-health-tech move, but it asks two things joined by 'and' (motivation and what should change), and the motivation question is fairly generic rather than about clinical experience or the non-coding angle.”

Plan a day in time slots
Passed: GPT-6 Luna, GPT-6 Sol, and GPT-6 Astra. Failed: Claude Sonnet 5.
Claude Sonnet 5 squeezed an hour of kitchen cleaning into 30 minutes before lunch, then pointed out the overlap instead of fixing it. The three GPT-6 models cleaned and called while the laundry ran, and everything fit.
- Why Claude Sonnet 5 failed: Graded 2.3 of 5 on average (lowest 2).
```
The grader: “The kitchen slot is labeled 1h but given only 30 minutes, and the reply itself admits it overlaps lunch (and miscalculates that as 25 min), so the listed schedule is broken; the laundry overlap idea is sensible, but the time math and the ~4.25h total are wrong, and a fix was suggested rather than ap”
```

Explain it like I'm five

Passed: GPT-6 Luna, GPT-6 Sol, and GPT-6 Astra. Failed: Claude Sonnet 5.

Why Claude Sonnet 5 failed: Graded 3.7 of 5 on average (lowest 2).

The grader: “The explanation is accurate and mostly simple, though 'immune system' and a few long sentences slip in, and it adds a fire-drill comparison (plus soldiers) on top of the wanted poster instead of using just one.”

Pros and cons with a verdict

Passed: GPT-6 Sol, GPT-6 Astra, and Claude Sonnet 5. Failed: GPT-6 Luna.

Why GPT-6 Luna failed: Graded 3.7 of 5 on average (lowest 3).

The grader: “Only two pros use the 8,000-mile and 3-year details, and none of the cons do (the last con even assumes keeping the car longer), the points are otherwise fair and accurate, and the one-sentence verdict leans toward leasing but hedges with conditions.”

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. Act as a career coach, one question at a time

Hard · 1 of 4 passed

Asking for questions one at a time turns a generic answer into a conversation about your case. The test is whether the model can hold back and ask just one.

The prompt: replace the parts in [brackets]

Act as a career coach. I'm [describe your situation]. Before you give any advice, ask me three questions, one at a time: ask only the first question now, then wait for my answer.

Open in llmwise

How we filled it in and checked the replies

Act as a career coach. I'm a nurse with six years on hospital wards, thinking about moving into health tech, but I don't code. Before you give any advice, ask me three questions, one at a time: ask only the first question now, then wait for my answer.

One question: Asks exactly one question now, not two or three at once.
Relevant: The question is specific to a ward nurse moving into health tech, not a generic career question.
Waits: Gives no advice yet, beyond at most a short line of acknowledgement.
Automatic rules: at most 90 words; ends with a question.

GPT-6 Luna
Failed
Graded 3.3 of 5 on average (lowest 2).
```
What do you enjoy most about your current nursing work, and what would you most like to change?
```
1 message on Pro · under $0.0001 to run · 1.1 s · 24 tokens out
GPT-6 Sol
Failed
Graded 3.7 of 5 on average (lowest 3).
```
What part of your nursing work do you most enjoy, and what would you like to do less of in your next role?
```
1 message on Pro · $0.0005 to run · 1.9 s · 29 tokens out

GPT-6 Astra

Failed

Graded 3.7 of 5 on average (lowest 3).

What’s motivating you to move from hospital nursing into health tech, and what would you like to be different in your next role?

1 message on Pro · $0.0027 to run · 1.5 s · 30 tokens out

Claude Sonnet 5

Passed

Graded 4.7 of 5 on average (lowest 4).

Question 1: What draws you to health tech specifically — is it a certain type of problem (clinical workflow, patient safety, data/analytics, etc.), a lifestyle change, or something else?

1 message on Pro · $0.0017 to run · 1.8 s · 63 tokens out

2. Explain it like I'm five

Everyday · 3 of 4 passed

A word limit and a ban on jargon make the model choose what matters, and one comparison keeps it from piling up metaphors.

The prompt: replace the parts in [brackets]

Explain [a topic] like I'm five. Under 100 words, no jargon, and use one everyday comparison.

Open in llmwise

How we filled it in and checked the replies

Explain how a vaccine teaches your body to fight germs like I'm five. Under 100 words, no jargon, and use one everyday comparison.

Five-year-old level: Simple words and short sentences a young child could follow.
Accurate: Gets the idea right: the vaccine shows the body a harmless look at a germ, so the body learns to recognise it and fights it off quickly later.
One comparison: Uses one everyday comparison that fits, not several.
Automatic rules: at most 100 words; doesn't use “antibod”; doesn't use “antigen”.

GPT-6 Luna

Passed

Graded 4.7 of 5 on average (lowest 4).

A vaccine is like a practice alarm for your body’s guards. It shows them a safe clue about a germ, so they can learn what to look for. If the real germ shows up later, your guards can spot it faster and fight it off.

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

GPT-6 Sol

Passed

Graded 4.7 of 5 on average (lowest 4).

A vaccine is like showing your body a picture of a sneaky germ. Your body practices spotting it and learns how to stop it. If the real germ shows up later, your body remembers what to do and can fight it faster.

1 message on Pro · $0.0006 to run · 1.6 s · 51 tokens out

GPT-6 Astra

Passed

Graded 4.7 of 5 on average (lowest 4).

A vaccine is like a practice game for your body’s guards. It safely shows them what a certain germ looks like, so they can learn to spot it and stop it.

If the real germ shows up later, your guards remember their practice and can fight back faster. That helps keep you from getting very sick.

1 message on Pro · $0.0035 to run · 19.5 s · 67 tokens out

Claude Sonnet 5

Failed

Graded 3.7 of 5 on average (lowest 2).

A vaccine is like a "wanted poster" for your body's germ-fighting police (your immune system). It shows a picture of a bad germ without the germ actually being dangerous. Your body looks at the poster, remembers the germ's face, and gets its tiny soldiers ready. So if that real germ ever shows up later, your body already knows exactly what it looks like and can catch it fast — before it makes you sick. It's like practicing a fire drill so you're ready if a real fire ever happens!

1 message on Pro · $0.0026 to run · 3.0 s · 159 tokens out

3. Three bullets and a TL;DR

Everyday · 4 of 4 passed

Fixing the number of bullets and the last line gives you a summary you can paste anywhere. The trap in this text: results come quarterly, though the opposition asked for monthly.

The prompt: replace the parts in [brackets]

Summarize the text below in exactly 3 bullet points, then one last line that starts with "TL;DR:". [paste the text]

Open in llmwise

How we filled it in and checked the replies

Summarize the text below in exactly 3 bullet points, then one last line that starts with "TL;DR:".

Dunmore City Council will trial a four-day working week for 300 of its office staff from 3 March. Staff will work 32 hours for the same pay, with offices open five days a week on a rota. The trial runs for six months and will be judged on three measures: how quickly residents' requests are answered, staff sick days, and the cost of agency cover. A similar trial in a neighbouring council last year cut sick days by 18% and saved £240,000 on agency staff, though response times for planning applications rose slightly. Unions welcomed the plan; the opposition group called it "an experiment with taxpayers' money" and asked for monthly public reports. The council has agreed to publish results every quarter. Bin collections, care workers and other front-line services are not part of the trial.

Key facts: Covers the trial (32 hours for the same pay, six months from 3 March), how it will be judged, and the reaction.
Accurate: States nothing the text doesn't say: results are published quarterly, not monthly.
TL;DR: The TL;DR line is one short sentence that stands on its own.
Automatic rules: exactly 3 bullet points; includes “TL;DR”.

GPT-6 Luna

Passed

Graded 4.7 of 5 on average (lowest 4).

- Dunmore City Council will trial a four-day week for 300 office staff from 3 March: 32 hours for the same pay, with offices remaining open five days via a rota.
- The six-month trial will track response times, staff sick days and agency-cover costs; front-line services are excluded.
- A neighbouring council’s trial reduced sick days by 18% and saved £240,000, but planning responses slowed slightly. Unions support the plan; critics want monthly reports, while Dunmore will publish results quarterly.

TL;DR: A six-month four-day-week trial will test effects on service, staff absence and costs, with quarterly public updates.

1 message on Pro · $0.0001 to run · 2.1 s · 159 tokens out

GPT-6 Sol

Passed

Graded 5.0 of 5 on average (lowest 5).

- From 3 March, Dunmore City Council will trial a four-day, 32-hour week for 300 office staff at full pay. Offices will remain open five days a week on a rota; front-line services are excluded.
- The six-month trial will measure response times to residents’ requests, staff sick days and agency-cover costs. Results will be published quarterly.
- Unions support the plan, while the opposition wants monthly reports. A neighbouring council’s trial reduced sick days and agency costs, but planning response times rose slightly.

TL;DR: Dunmore will test a four-day week for office staff while tracking service speed, absences and costs.

1 message on Pro · $0.0012 to run · 1.8 s · 137 tokens out

GPT-6 Astra

Passed

Graded 5.0 of 5 on average (lowest 5).

- Dunmore City Council will trial a four-day, 32-hour week on unchanged pay for 300 office staff from 3 March for six months. Offices will remain open five days through rotas; front-line services are excluded.
- Success will be assessed through residents’ request response times, staff sick days and agency cover costs, with quarterly results published. A neighbouring council’s trial reduced sick days by 18% and saved £240,000, but planning responses slowed slightly.
- Unions welcomed the trial; the opposition criticised it as an experiment with taxpayers’ money and requested monthly public reports.

TL;DR: Dunmore will test shorter office-staff hours without cutting pay, tracking service performance, sickness and agency costs.

1 message on Pro · $0.0064 to run · 3.2 s · 151 tokens out

Claude Sonnet 5

Passed

Graded 4.0 of 5 on average (lowest 3).

- Dunmore City Council will trial a four-day week for 300 office staff starting March 3, with 32-hour weeks at full pay while offices stay open five days via rota, for six months.
- Success will be measured by response times to residents, staff sick days, and agency cover costs — metrics inspired by a neighboring council's trial that cut sick days 18% and saved £240,000, though planning response times slipped slightly.
- Unions support the trial, while opposition members criticized it as risky spending and pushed for reporting; the council agreed to publish results quarterly instead of monthly, and front-line services like bins and care work are excluded.

TL;DR: Dunmore Council is piloting a four-day week for office staff with quarterly public reporting, drawing mixed reactions but backed by promising results from a similar nearby trial.

1 message on Pro · $0.0043 to run · 5.0 s · 278 tokens out

4. Ten ideas within a budget

Everyday · 4 of 4 passed

A number, a budget and a way to mark the cheapest turn a brainstorm into a list you can act on.

The prompt: replace the parts in [brackets]

Give me 10 ideas for [what you're planning], with a total budget of [your budget]. Number them, one line each, and put a star (*) after the three cheapest.

Open in llmwise

How we filled it in and checked the replies

Give me 10 ideas for a birthday party for a 9-year-old who loves space, with a total budget of $100. Number them, one line each, and put a star (*) after the three cheapest.

On theme: Every idea suits a 9-year-old's space-themed party.
Within budget: Each idea is doable within $100 in total; none obviously needs more.
Three starred: Exactly three ideas are starred, and they're plausibly the cheapest.
Automatic rules: exactly 10 bullet points.

GPT-6 Luna

Passed

Graded 5.0 of 5 on average (lowest 5).

1. Backyard stargazing with borrowed binoculars and homemade constellation cards — $0*
2. DIY space-themed party games using household supplies — $5*
3. Make-and-launch paper rockets in the yard — $8*
4. Glow-in-the-dark space dance party with a playlist and glow sticks — $12
5. Astronaut training obstacle course using cushions and chairs — $15
6. Moon-surface sensory bin with flour and toy rovers — $18
7. Space-themed movie night with popcorn and homemade tickets — $20
8. Build-your-own alien masks with craft-store supplies — $25
9. Planet-making craft party with foam balls and paint — $30
10. Make mini “solar system” cupcakes together — $40

1 message on Pro · $0.0001 to run · 2.6 s · 159 tokens out

GPT-6 Sol

Passed

Graded 5.0 of 5 on average (lowest 5).

1. Backyard stargazing with a constellation bingo sheet and homemade cake — about $40 total.*
2. Space scavenger hunt with hand-drawn clues and small prizes — about $45 total.*
3. Space movie night with popcorn, glow sticks, and homemade cake — about $50 total.*
4. Make-your-own alien masks with craft supplies and themed snacks — about $55 total.
5. Moon-crater experiments using flour, cocoa, and marbles — about $60 total.
6. Build cardboard-box spaceships and hold a pretend launch — about $65 total.
7. Paint paper-plate planets and create a solar-system mural — about $70 total.
8. Launch DIY straw rockets and award mission badges — about $75 total.
9. Host astronaut training games with an obstacle course and medals — about $85 total.
10. Throw a glow-in-the-dark galaxy dance party with decorations and snacks — about $95 total.

1 message on Pro · $0.0025 to run · 6.2 s · 429 tokens out

GPT-6 Astra

Passed

Graded 5.0 of 5 on average (lowest 5).

1. Astronaut training: Build an obstacle course with chairs, cushions, and “moon jumps”—$0 using household items. *
2. Alien freeze dance: Play space-themed music and freeze in silly alien poses—$0. *
3. Mission Control says: Play a space version of Simon Says—$0. *
4. Moon-rock hunt: Hide foil-wrapped stones and give kids clues to find them—$5.
5. Galaxy decorations: Create a starry backdrop with black paper, chalk, and balloons—$8.
6. Rocket workshop: Build cardboard-tube rockets with colored paper and stickers—$10.
7. Space explorer favors: Give each guest glow sticks and space stickers—$12.
8. Moon-crater cake: Decorate a homemade cake with cookie craters—$15.
9. Planet pizza dinner: Serve cheese pizzas with optional “asteroid” vegetable toppings—$20.
10. Astronaut snack station: Offer popcorn “moon rocks,” fruit “planets,” and lemonade “rocket fuel”—$25; estimated total for eight kids: $95.

1 message on Pro · $0.0165 to run · 11.2 s · 584 tokens out

Claude Sonnet 5

Passed

Graded 4.3 of 5 on average (lowest 4).

1. DIY "planet" balloons hung from ceiling with black streamers *
2. Print free NASA photos as decorations *
3. Homemade "astronaut" cardboard box helmets for kids *
4. Glow-in-the-dark star stickers on ceiling/walls
5. Space-themed cake (order or bake with black icing and star sprinkles)
6. Constellation flashlight game in a dark room using a paper cup with holes
7. DIY rocket craft station using toilet paper rolls and paint
8. Backyard "moon landing" scavenger hunt with clues
9. Rent or buy a cheap telescope/binoculars for stargazing at dusk
10. Space-themed party favor bags with small toys (astronaut figures, glow bracelets)

1 message on Pro · $0.0038 to run · 4.7 s · 276 tokens out

5. Meeting notes into a table, no guessing

Hard · 4 of 4 passed

"Leave it empty, don't guess" is the line that stops a model filling gaps with plausible names and dates.

The prompt: replace the parts in [brackets]

Turn these meeting notes into a table with the columns Task, Owner and Due date. If the notes don't give an owner or a date, leave the cell empty: don't guess. [paste your notes]

Open in llmwise

How we filled it in and checked the replies

Turn these meeting notes into a table with the columns Task, Owner and Due date. If the notes don't give an owner or a date, leave the cell empty: don't guess.

Notes, product sync, 6 Oct:
- Priya will update the pricing page before the launch on 20 Oct.
- Someone needs to check the refund emails still work.
- Marco to send the beta survey by Friday 10 Oct.
- Legal review of the new terms: Dana, no date yet.
- We should think about a webinar in November.

All tasks: Has a row for each of the five tasks.
No guessing: Leaves the owner empty for the refund emails and the webinar, and the date empty for the refund emails and the legal review; invents no owner or date.
Right values: Priya by 20 Oct (or before it), Marco by 10 Oct, Dana for the legal review.
Automatic rules: includes “Priya”; includes “Marco”; includes “Dana”.

GPT-6 Luna

Passed

Graded 5.0 of 5 on average (lowest 5).

| Task | Owner | Due date |
|---|---|---|
| Update the pricing page | Priya | Before 20 Oct |
| Check that the refund emails still work |  |  |
| Send the beta survey | Marco | 10 Oct |
| Review the new terms | Dana |  |
| Consider a webinar |  | November |

1 message on Pro · under $0.0001 to run · 1.2 s · 77 tokens out

GPT-6 Sol

Passed

Graded 5.0 of 5 on average (lowest 5).

| Task | Owner | Due date |
|---|---|---|
| Update the pricing page | Priya | Before 20 Oct |
| Check that the refund emails still work | | |
| Send the beta survey | Marco | Friday 10 Oct |
| Review the new terms | Dana | |
| Consider a webinar in November | | |

1 message on Pro · $0.0013 to run · 2.9 s · 162 tokens out

GPT-6 Astra

Passed

Graded 5.0 of 5 on average (lowest 5).

| Task | Owner | Due date |
|---|---|---|
| Update the pricing page | Priya | Before 20 Oct |
| Check that refund emails still work | | |
| Send the beta survey | Marco | Friday 10 Oct |
| Review the new terms | Dana | |
| Consider a webinar in November | | |

1 message on Pro · $0.0059 to run · 3.1 s · 145 tokens out

Claude Sonnet 5

Passed

Graded 5.0 of 5 on average (lowest 5).

| Task | Owner | Due date |
|---|---|---|
| Update the pricing page before launch | Priya | 20 Oct |
| Check the refund emails still work | | |
| Send the beta survey | Marco | Fri 10 Oct |
| Legal review of the new terms | Dana | |
| Think about a webinar in November | | |

1 message on Pro · $0.0025 to run · 2.1 s · 122 tokens out

6. Firm but polite, every fact kept

Everyday · 4 of 4 passed

"Keep every fact" matters most in an angry message: the dates, the missed visit and the deadline are what make it firm.

The prompt: replace the parts in [brackets]

Rewrite this message to [who it's for] so it's firm but polite. Keep every fact and date. Under 120 words. Show only the rewrite. [paste your message]

Open in llmwise

How we filled it in and checked the replies

Rewrite this message to my landlord so it's firm but polite. Keep every fact and date. Under 120 words. Show only the rewrite.

hey this is the THIRD time im asking. the boiler broke on 2 october and its now 16 october and we have had no hot water for two weeks. you said someone would come on the 9th and nobody came. i pay £1,150 a month and i want it fixed by friday 18th or im going to contact the council about it.

Firm but polite: Calm and courteous, but leaves no doubt it must be fixed.
Every fact: Keeps the dates (2, 9, 16 and 18 October), the missed visit, the rent and the council deadline.
Clear ask: Says clearly what is wanted and by when.
Automatic rules: at most 120 words; includes “2 October”; includes “9”; includes “18”; includes “1,150”; includes “council”.

GPT-6 Luna

Passed

Graded 4.7 of 5 on average (lowest 4).

This is my third time raising this issue. The boiler broke on 2 October, and as of 16 October, we have been without hot water for two weeks. You said someone would come on the 9th, but nobody came. I pay £1,150 a month, and I expect the boiler to be fixed by Friday 18th. If it isn’t, I will contact the council.

1 message on Pro · under $0.0001 to run · 1.6 s · 87 tokens out

GPT-6 Sol

Passed

Graded 5.0 of 5 on average (lowest 5).

Dear Landlord,

This is my third request about the boiler, which broke on 2 October. It is now 16 October, and we have had no hot water for two weeks. You said someone would come on the 9th, but nobody arrived. I pay £1,150 a month, and I ask that the boiler be fixed by Friday 18th. If it is not, I will contact the council.

Kind regards,

1 message on Pro · $0.0009 to run · 1.5 s · 94 tokens out

GPT-6 Astra

Passed

Graded 5.0 of 5 on average (lowest 5).

This is my third request regarding the broken boiler. It broke on 2 October, and today is 16 October, meaning we have been without hot water for two weeks. You said someone would attend on 9 October, but nobody came.

I pay £1,150 per month in rent and expect this issue to be resolved promptly. Please ensure the boiler is repaired by Friday 18 October. If it is not fixed by then, I will contact the council.

Please confirm when the repair will take place.

1 message on Pro · $0.0049 to run · 2.0 s · 107 tokens out

Claude Sonnet 5

Passed

Graded 4.3 of 5 on average (lowest 4).

Hi [Landlord's name],

This is the third time I'm reaching out about this. The boiler broke on 2 October, and it's now 16 October — we've been without hot water for two weeks. You said someone would come on the 9th, but no one attended.

I pay £1,150 a month in rent and would appreciate this being resolved by Friday 18th. If it isn't fixed by then, I'll need to contact the council for further advice.

Thank you for your prompt attention to this.

1 message on Pro · $0.0027 to run · 2.7 s · 154 tokens out

7. Pros and cons with a verdict

Everyday · 3 of 4 passed

Giving your own numbers (mileage, how long you keep a car) is what makes the pros and cons yours rather than a generic list.

The prompt: replace the parts in [brackets]

Give me the pros and cons of [your decision]: exactly 4 pros and 4 cons as bullet points, then a one-sentence verdict that isn't a bullet.

Open in llmwise

How we filled it in and checked the replies

Give me the pros and cons of leasing a new car instead of buying one, for someone who drives about 8,000 miles a year and changes cars every 3 years: exactly 4 pros and 4 cons as bullet points, then a one-sentence verdict that isn't a bullet.

Specific: The points use the low mileage and the 3-year habit, not just generic leasing facts.
Balanced: The pros and cons are fair and accurate.
Verdict: Ends with a one-sentence verdict that leans one way and follows from the points.
Automatic rules: exactly 8 bullet points.

GPT-6 Luna

Failed

Graded 3.7 of 5 on average (lowest 3).

- **Pro:** Your 8,000 miles a year is well below many lease mileage limits, so you’re less likely to face excess-mileage charges.
- **Pro:** Leasing makes it easy to drive a new car every three years, without selling or trading in an older one.
- **Pro:** Monthly payments are often lower than financing the same new car, freeing up cash during the lease.
- **Pro:** You’re typically covered by the manufacturer’s warranty for much of the lease, which can reduce repair costs.
- **Con:** You build no ownership equity, and payments continue if you lease another car after three years.
- **Con:** The lease may charge for excess wear, damage, or mileage beyond its contract limits.
- **Con:** You’re bound by the lease terms; ending early can be costly, and customizing the car may be restricted.
- **Con:** Buying can cost less over the long run if you keep a car beyond three years or find a strong resale value.

**Verdict:** Given your low mileage and preference to change cars every three years, leasing may suit you well if you’re comfortable with ongoing payments and the contract’s restrictions.

1 message on Pro · $0.0001 to run · 3.0 s · 241 tokens out

GPT-6 Sol

Passed

Graded 4.3 of 5 on average (lowest 4).

**Pros**
- Your 8,000 miles a year will likely fit within a standard lease’s mileage allowance.
- A three-year lease matches how often you change cars, so you can return the car rather than sell or trade it.
- Monthly payments are often lower than payments on a loan for the same new car.
- The car will typically remain under its factory warranty throughout the lease.

**Cons**
- You won’t build equity or own the car at the end unless you buy it.
- You may owe charges for excess wear or damage when you return it.
- Mileage limits still matter if your driving increases unexpectedly.
- Lease fees and repeated payments can cost more over time than buying and keeping a car longer.

For your low mileage and three-year replacement cycle, leasing is a strong fit if you value convenience and predictable costs more than ownership.

1 message on Pro · $0.0013 to run · 2.4 s · 175 tokens out

GPT-6 Astra

Passed

Graded 4.7 of 5 on average (lowest 4).

**Pros**
- **Usually lower monthly payments:** Leasing generally costs less per month than financing the same new car.
- **Fits your replacement cycle:** A typical 36-month lease matches your habit of changing cars every three years.
- **Low-mileage options:** At 8,000 miles annually, you may qualify for a cheaper low-mileage lease.
- **Warranty coverage:** Most repairs should remain covered by the manufacturer’s warranty during a three-year lease.

**Cons**
- **No ownership equity:** Your payments don’t build ownership, and you won’t receive resale proceeds when you return the car.
- **Limited benefit from low mileage:** Buying lets you capture any higher resale value from driving less; leasing generally provides no refund for unused miles.
- **Potential extra charges:** Excess wear, mileage overages, and lease-end fees can increase your total cost.
- **Less flexibility:** Ending a lease early can be expensive, and modifications are typically restricted.

Leasing is a strong fit for your low mileage and three-year replacement cycle, but compare total three-year costs—including fees and a purchased car’s likely resale value—before deciding.

1 message on Pro · $0.0078 to run · 5.1 s · 233 tokens out

Claude Sonnet 5

Passed

Graded 4.0 of 5 on average (lowest 4).

# Leasing a New Car: Pros and Cons

## Pros
- **Low mileage fits perfectly**: At 8,000 miles/year, you'll likely stay well under standard lease limits (usually 10,000–12,000/year), avoiding overage fees entirely.
- **Lower monthly payments**: Leasing typically costs less per month than financing a purchase, since you're only paying for depreciation over the lease term, not the full vehicle value.
- **Always under warranty**: With a 3-year cycle, your car stays covered by the manufacturer's warranty the entire time you own it, minimizing repair costs.
- **Latest tech and safety features**: Frequent swaps mean you consistently get newer models with updated infotainment, safety systems, and fuel/EV efficiency improvements.

## Cons
- **No equity or ownership**: You're paying for use, not building any asset—at lease-end, you walk away with nothing to show for the payments.
- **Long-term cost is higher**: If you plan to keep cycling every 3 years indefinitely, leasing almost always costs more over time than buying and driving a car longer.
- **Mileage overages still risk you**: While unlikely at 8,000 miles/year, unexpected life changes (new job, road trips) could push you over limits and trigger costly per-mile penalties.
- **Customization and wear restrictions**: Leases often penalize excess wear-and-tear or modifications, limiting flexibility compared to owning outright.

Given your low mileage and preference for a new car every three years, leasing is likely the more cost-effective and hassle-free choice—provided you're comfortable never building equity in the vehicle.

1 message on Pro · $0.0066 to run · 8.2 s · 549 tokens out

8. Plan a day in time slots

Hard · 3 of 4 passed

Hours, fixed breaks and which tasks can overlap are the details a schedule needs; without them you get a vague list.

The prompt: replace the parts in [brackets]

Plan my [day] as a list of time slots. I'm free [your hours], with lunch [lunch]. I need to fit in: [your tasks]. Say if anything doesn't fit.

Open in llmwise

How we filled it in and checked the replies

Plan my Saturday as a list of time slots. I'm free from 9:00 to 17:00, with lunch 12:30–13:30. I need to fit in: laundry (2 hours, but I can do other things while it runs), a grocery run (1 hour), the gym (1.5 hours), calling my mom (30 minutes), and cleaning the kitchen (1 hour). Say if anything doesn't fit.

Everything fits: Every task has a slot between 9:00 and 17:00, and nothing overlaps lunch.
Sensible: Uses the laundry's running time for another task, and no two hands-on tasks overlap; the times add up.
Clear: An easy-to-follow list of time slots.

GPT-6 Luna

Passed

Graded 5.0 of 5 on average (lowest 5).

- **9:00–11:00** — Laundry (running in the background)
- **9:00–10:00** — Clean the kitchen
- **10:00–10:30** — Call your mom
- **11:00–12:00** — Grocery run
- **12:30–13:30** — Lunch
- **13:30–15:00** — Gym
- **15:00–17:00** — Free time / buffer

Everything fits.

1 message on Pro · $0.0001 to run · 4.6 s · 412 tokens out

GPT-6 Sol

Passed

Graded 5.0 of 5 on average (lowest 5).

- **9:00–11:00** — Laundry runs
  - **9:00–10:00** — Clean the kitchen
  - **10:00–10:30** — Call your mom
  - **10:30–11:00** — Finish or fold laundry
- **11:00–12:00** — Grocery run
- **12:00–12:30** — Free time
- **12:30–13:30** — Lunch
- **13:30–15:00** — Gym
- **15:00–17:00** — Free time

Everything fits.

1 message on Pro · $0.0022 to run · 3.6 s · 362 tokens out

GPT-6 Astra

Passed

Graded 5.0 of 5 on average (lowest 5).

- **9:00–10:00:** Start laundry, then clean the kitchen. Laundry runs until 11:00.
- **10:00–10:30:** Call your mom while laundry continues.
- **10:30–11:00:** Take a break; finish up laundry at 11:00.
- **11:00–12:00:** Grocery run.
- **12:00–12:30:** Put groceries away and prepare lunch.
- **12:30–13:30:** Lunch.
- **13:30–15:00:** Gym.
- **15:00–17:00:** Free time, with room for travel or any tasks that run over.

**Everything fits**, with laundry overlapping kitchen cleaning and your call.

1 message on Pro · $0.0106 to run · 5.7 s · 336 tokens out

Claude Sonnet 5

Failed

Graded 2.3 of 5 on average (lowest 2).

Everything fits comfortably. Here's a plan that uses the laundry's "downtime" efficiently:

**9:00 – 9:10** — Start laundry (load washer)
**9:10 – 10:10** — Grocery run (1h)
**10:10 – 10:15** — Switch laundry to dryer
**10:15 – 11:45** — Gym (1.5h)
**11:45 – 11:55** — Fold/put away laundry
**11:55 – 12:25** — Clean kitchen (1h — squeezes to 12:30, see below)
**12:30 – 13:30** — Lunch
**13:30 – 14:00** — Call mom (30 min)
**14:00 – 17:00** — Free time (3 hours to spare!)

Note: kitchen cleaning is scheduled 11:55–12:55 in reality if you take the full hour, which overlaps lunch by 25 min — easiest fix is to just shift lunch slightly or trim kitchen cleaning by 5 min that day. Alternatively, swap order: do kitchen cleaning *after* lunch and call mom before lunch — either works since you have 3 free hours in the afternoon as buffer.

Total active task time: ~4.25 hours out of your 7 available (excluding lunch) — nothing gets cut, and you end up with a relaxed afternoon.

1 message on Pro · $0.0158 to run · 14.1 s · 1,457 tokens out

9. A skeptical investor, no praise

Hard · 4 of 4 passed

Asking for a skeptic and banning praise gets past the polite encouragement models default to.

The prompt: replace the parts in [brackets]

Act as a skeptical investor. Here's my business idea: [your idea]. Give me the 3 biggest risks as bullet points, one sentence each, then one question I must answer before I start. Don't praise the idea.

Open in llmwise

How we filled it in and checked the replies

Act as a skeptical investor. Here's my business idea: a subscription box of houseplants for people in city apartments, $29 a month, shipped nationwide. Give me the 3 biggest risks as bullet points, one sentence each, then one question I must answer before I start. Don't praise the idea.

Real risks: Names serious, specific risks (live plants in shipping, plants dying, churn, margins), not generic ones.
No praise: No compliments or softening about the idea.
Sharp question: Ends with one question that goes to the heart of whether it works.
Automatic rules: exactly 3 bullet points; ends with a question; doesn't use “great idea”; doesn't use “love this”; doesn't use “exciting”.

GPT-6 Luna

Passed

Graded 4.7 of 5 on average (lowest 4).

- **Unit economics may not work:** At $29 per month, plant costs, packaging, shipping, damage replacements, and customer acquisition could leave little or no contribution margin.
- **Retention may be weak:** Customers may run out of space, struggle to care for plants, or lose interest after the novelty wears off, making recurring revenue hard to sustain.
- **Nationwide fulfillment is operationally risky:** Plants are perishable and sensitive to weather and handling, so transit damage, seasonal constraints, and replacement costs could erode margins and trust.

**Before you start:** What is your expected contribution margin per box after shipping, packaging, damage, and customer acquisition costs?

1 message on Pro · $0.0001 to run · 2.0 s · 138 tokens out

GPT-6 Sol

Passed

Graded 5.0 of 5 on average (lowest 5).

- **Unit economics:** At $29 a month, plant sourcing, packaging, nationwide shipping, replacements, and customer acquisition could leave little or no margin.
- **Churn:** Apartment dwellers may run out of space or lose interest after a few deliveries, making it hard to recover acquisition costs.
- **Shipping losses:** Heat, cold, delays, and rough handling can kill or damage plants, driving refunds and cancellations.

**Question:** After shipping, replacements, and acquisition costs, how many months must the average customer stay subscribed for you to make a profit?

1 message on Pro · $0.0010 to run · 1.9 s · 117 tokens out

GPT-6 Astra

Passed

Graded 5.0 of 5 on average (lowest 5).

- **Thin or negative margins:** At $29 a month, plants, protective packaging, nationwide shipping, weather-related losses, and replacements could consume your revenue before marketing and overhead.
- **High churn:** Small apartments have limited space, so even satisfied customers may cancel once they have enough plants, undermining subscription lifetime value.
- **Expensive customer acquisition:** You compete with local nurseries and established online sellers, and acquiring subscribers could cost more than their total contribution profit.

**Before you start:** What evidence shows customers will stay subscribed long enough for contribution profit—after fulfillment, losses, and replacements—to exceed acquisition cost?

1 message on Pro · $0.0052 to run · 3.5 s · 129 tokens out

Claude Sonnet 5

Passed

Graded 5.0 of 5 on average (lowest 5).

- **Shipping mortality**: Live plants shipped nationwide face temperature extremes, transit delays, and rough handling, meaning high breakage/death rates that eat margins and trigger refund demands.
- **Weak retention**: Unlike consumables, plants don't get "used up"—once a subscriber has 6-8 plants filling their apartment, they have no reason to keep paying $29/month, so churn will likely spike fast.
- **Thin unit economics**: Between plant cost, pot/packaging, insulated shipping materials, and return/refund losses, $29 leaves little room for customer acquisition cost, making profitability fragile unless you nail retention or upsells.

**Question you must answer**: What is your actual customer lifetime value once you factor in realistic churn (likely under 6 months) versus your fully-loaded cost per shipment (plant + packaging + shipping + damage/refund rate)?

1 message on Pro · $0.0041 to run · 4.1 s · 300 tokens out

10. Write it in my style

Everyday · 4 of 4 passed

Two real samples say more about your voice than any description of it; the length is the part models most often miss.

The prompt: replace the parts in [brackets]

Here are two emails I wrote: [paste two of your emails] Write a new email in the same style asking [what you need]. Match my length, tone and sign-off.

Open in llmwise

How we filled it in and checked the replies

Here are two emails I wrote:

Email 1:
Hi all,
Quick one: the kitchen fridge gets cleared out Friday at 4. Anything unlabelled goes.
Thanks!
Jo

Email 2:
Hi all,
Reminder that Monday's stand-up moves to 9:30 this week (room B). Same agenda.
Thanks!
Jo

Write a new email in the same style asking my team to submit their September expenses by Thursday. Match my length, tone and sign-off.

Same style: Short and direct like the samples, with the same opener.
Clear ask: Asks for September expenses by Thursday.
Sign-off: Signs off "Thanks!" then "Jo", as the samples do.
Automatic rules: at most 45 words; includes “Thanks!”; includes “Jo”; includes “Thursday”.

GPT-6 Luna
Passed
Graded 4.7 of 5 on average (lowest 4).
```
Hi all,
Please submit your September expenses by Thursday.
Thanks!
Jo
```
1 message on Pro · under $0.0001 to run · 1.1 s · 18 tokens out
GPT-6 Sol
Passed
Graded 5.0 of 5 on average (lowest 5).
```
Hi all,
Quick reminder: please submit your September expenses by Thursday.
Thanks!
Jo
```
1 message on Pro · $0.0005 to run · 0.9 s · 21 tokens out
GPT-6 Astra
Passed
Graded 5.0 of 5 on average (lowest 5).
```
Hi all,
Quick reminder: please submit your September expenses by Thursday.
Thanks!
Jo
```
1 message on Pro · $0.0027 to run · 1.6 s · 21 tokens out
Claude Sonnet 5
Passed
Graded 4.7 of 5 on average (lowest 4).
```
Hi all,
Quick one: September expenses are due Thursday.
Thanks!
Jo
```
1 message on Pro · $0.0016 to run · 1.7 s · 34 tokens out

Getting more from these prompts

Say who it's for and what you'll do with the answer: "for my landlord", "to paste into a spreadsheet".
Give the numbers, names and dates yourself. A model that has to guess them will, confidently.
Ask for a format you can check: a number of bullets, a word limit, a table with named columns.
When you want a conversation, say "one question at a time" and "wait for my answer"; most models ask everything at once otherwise.

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

Were these replies made in ChatGPT?

No. llmwise isn't ChatGPT and doesn't offer it: it has OpenAI's GPT-6 Astra, Sol and Luna, among models from other AI companies, and these replies come from those models run through llmwise. The prompts are plain text, so they work the same way pasted into ChatGPT.

Do these prompts work on every GPT model?

Most did on every model; the ones that split the models are listed near the top, with the check each reply failed. In our runs GPT-6 Sol passed the most: 9 of 10.

Can I try these prompts for free?

Yes: copy one, fill in the parts in [brackets], and send it. 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.

Start free See pricing