What the tech
Subscribe

Search

/ opens search from anywhere. Esc closes it.

Get the newsletter

AI Compare · Test kits

Test AI tools on your own work

Run the same test in each AI, in your own account, and score the result against the same checks.

Prices for United States, before sales tax Data release 2026.10.16-3, Oct 16, 2026 Rules 2.1.0

A test kit is the fairest way to choose: the same files, the same prompt and the same checks for every AI you try. You run it in your own account, so the result reflects your plan, your settings and your limits. This site never runs anything for you and never sees what you paste.

Research & Search 4 kits

  • Summarize a report with traceable facts

    A one-page quarterly report with ten decision facts, a figure that excludes a new store, and a price rise that has not been decided. A good summary keeps every fact, says where it came from, and adds nothing.

    Fixed files 9 checks · version 1.0.0

  • Research a question with sources that support the answer

    A frozen packet of four dated documents about a bus fare cut, including a data note that weakens the headline figure and a blog post with an unsupported claim. This is the controlled version; a live-web test needs its own date window.

    Fixed files 5 checks · version 1.0.0

  • Make sense of documents that disagree

    Four dated documents about a city scooter pilot: a city report says injuries fell, a hospital note and a newspaper say they rose, and the operator claims none were serious. The numbers measure different things over different months.

    Fixed files 8 checks · version 1.0.0

  • Research a current question within a date window

    An open-web question with an answer that changes every month: the latest US inflation figure from its official source. You check the answer yourself on the official site, on the same day, and record when and how you ran it.

    Open web 7 checks · version 1.0.0

Writing & Content 3 kits

  • Edit an email without losing a detail

    A rushed internal email with a moved meeting, a budget cap, a deadline and a confidentiality request. The rewrite must keep every detail and promise nothing new.

    Fixed files 9 checks · version 1.0.0

  • Edit an article without changing its facts

    A 370-word local news draft with eleven figures, six dates, a quotation, an objection and two hedges. Cut it to about 300 words and check that every fact survives and nothing uncertain becomes certain.

    Fixed files 9 checks · version 1.0.0

  • Turn project notes into a report that agrees with itself

    Status notes for a clinic expansion: a budget to add up, a supplier delay, decisions taken and questions still open. The report must get the totals right, keep the decided apart from the undecided, and invent no opening date.

    Fixed files 9 checks · version 1.0.0

Images & Design 4 kits

  • Make a picture that follows a detailed brief

    One paragraph with countable details: a tram with exactly three passengers, a woman in a green coat with a white dog, four houses with red roofs, an evening sky and no text. Score the first image against each detail.

    Fixed files 8 checks · version 1.0.0

  • Put exact text into a graphic

    A library poster with three lines of text that must be spelled exactly, a middle dot and a time range included, and no other text at all. Zoom in and compare letter by letter.

    Fixed files 7 checks · version 1.0.0

  • Change one thing in a picture and nothing else

    A flat illustration of a desk with seven distinct objects. The task is to turn the red mug blue. Everything else must stay exactly where and how it was, so you can compare the two pictures side by side.

    Fixed files 6 checks · version 1.0.0

  • Design a logo with an exact name

    A two-color logo for a fictional bakery: one wheat stalk and the name in one line, flat, nothing else. Checks the spelling, the restraint and whether it still reads small, and records whether you can get a vector (SVG) file.

    Fixed files 7 checks · version 1.0.0

Coding & Development 3 kits

  • Diagnose and repair a small program

    A twenty-line Python function with three real bugs and six unit tests. Run the tests on your own computer: this site never runs code.

    Fixed files 5 checks · version 1.0.0

  • Implement a small specification

    A written specification for a Python function, seven visible tests and five held-out tests you keep to yourself. Held-out tests show whether the code follows the specification or only the tests it was shown.

    Fixed files Held-out tests 5 checks · version 1.0.0

  • Refactor code without changing what it does

    A working but deeply nested shipping-cost function, seven tests to keep passing and nine held-out tests that pin rounding, error order and edge cases. A refactor that changes behaviour fails, however clean it looks.

    Fixed files Held-out tests 6 checks · version 1.0.0

Video & Animation 1 kit

  • Make a short clip that follows a brief

    A five-second shot with one camera move and one small event: a yellow leaf falls into a bicycle basket. Checks that the clip follows the brief and that the bicycle stays the same bicycle from the first frame to the last.

    Fixed files 7 checks · version 1.0.0

Data & Analytics 1 kit

  • Analyze a spreadsheet and get the numbers right

    Forty orders with cancelled and refunded rows, two missing prices and a correction row with negative units. The answers are fixed, so you can check them exactly.

    Fixed files 6 checks · version 1.0.0

Audio & Voice 2 kits

  • Make a song that sings your lyrics exactly

    Original lyrics (two verses and a chorus sung twice) with a style, tempo and length to follow. Checks that every line is sung, in order and understandably, and that the song matches the brief.

    Fixed files 8 checks · version 1.0.0

  • Read a script with names, numbers and abbreviations

    A short community update full of things text to speech gets wrong: a double-barrelled name, Dr. and St., times, money with cents, a percentage, a web address and a road number. Score each item by ear.

    Fixed files 8 checks · version 1.0.0

No result here is paid for. Products cannot buy a better position, and no affiliate link is used in any comparison or recommendation.

Product names and logos are trademarks of their owners (ChatGPT and the OpenAI logo of OpenAI, Claude of Anthropic, Gemini of Google, and so on). Logos are shown only to identify each product and do not imply endorsement.

Found an error? Tell us at our contact page and we will correct it in the change log.