⚡ Interactive benchmark✓ Scored against known answers🎯 New test type: Box

🧠 Decision API Showdown

Give the same text or image to 16 decision AIs and ask them to make a call — yes or no, pick one, or rate it. See who gets it right, who's fastest, and who's cheapest.

Engines16 live
Scenarios14 real-world
Hot takes4 unscored
Fair use5 tests/min
New test typeBoxThe AI points to where something is, not just yes or no.
New: Game ArenaWatch two decision AIs play Connect Four or Tic-tac-toe against each other, one decision per move.Watch a match →
1

Pick a scenario

A short, realistic piece of text and a few questions about it.

🎯 Box (8) →
2

What the AI reads, and what it's asked

Every contestant receives exactly this text and these questions.

hi-mum-scam.json
// Every AI reads exactly this
Text message
  From: Unknown number
  Message: Hi mum, I dropped my phone in the sink so this is my new number 🙈 can you help me pay a bill today? I'll pay you back tomorrow, it's urgent x
Length: 38 words189 charsFormat: JSON

Only this built-in text goes to the AI providers. Your own text is never sent from this page.

Task 1: Yes / no decisionkey: is_scam

Is the message in text_message likely a scam?

Value: yes / nowith a probability for each
✓ Correct answer: Yes
Task 2: Pick onekey: tactic

Which trick does the message in text_message mainly rely on?

impersonation— Pretending to be a family member or friend
urgency— Pressure to act immediately
reward— Promise of a prize, refund or money
none— No trick, it looks genuine
✓ Correct answer: impersonation
Task 3: Rating scalekey: risk

How risky would it be to send money in reply to text_message?

0Safe
1Low risk
2Risky
3Dangerous
✓ Correct answer: 3 · Dangerous
3

Choose the contestants and run

Toggle the AIs to include. They all get the request at the same time.

16 of 16 selected
Fair-use limit: 5 tests per minute
4

Results

Accuracy, speed and cost side by side.

🏁Pick a scenario and press Run showdown to see the seven AIs go head to head.
Have a use case?

Tell us the decision you'd hand to an AI. It might become the next scenario.

Submit an idea
Get updates

New AI contestants, fresh scenarios and upgrades, straight to your inbox.

Keep me posted
Like these tests?

Every run calls 16 paid AI APIs. Chip in to keep the tests free for everyone.