Get your score before the GPT-5.6 Luna benchmark is published.
Discover your profile now. It can be compared as soon as a documented run sets the target.
Start the test →The human × ChatGPT challenge
Can you beat ChatGPT?
ChatGPT has taken the M.I. challenge. Now it is your turn: 13–26 minutes, five dimensions and one score to compare. You do not need to lead in every dimension to reveal a human advantage.
HUMANS × CHATGPT
The next round against ChatGPT is being prepared.
Take the test now. As soon as this version has a published GPT-5.6 Luna target, your score will be ready for the head-to-head.
TARGET BEING PREPAREDM.I. · 10–190
The next score to beat will appear here.
No target score has been documented for this version yet. Get a head start: your first attempt will be ready for comparison.
Get your team ready for the next round.
Build your team’s overall profile now, then compare it with GPT-5.6 Luna as soon as a benchmark for this version is published.
Explore bespoke assessments →How this comparison works
Experimental, non-standardised benchmark using a simulated reference, not a human norm. Human and API runs use the same 29 scored task families and base formula, but presentation and timing differ. The human path also includes a chess challenge outside the five dimensions: solving both moves adds 5 points to the final score, shown separately and not offered in the GPT run. GPT-5.6 Luna completed the assessment through Codex with medium reasoning effort; M.I. then reviewed and published the result. The operator declares the model name and execution context; OpenAI does not authenticate either. This result does not measure ChatGPT as a whole or establish general human or AI superiority.
Beyond the total
The overall score is not the whole story.
You may trail ChatGPT overall and still outscore it in one or more dimensions. When a comparison is available, your result, share card and certificate highlight those advantages.
Outscoring the benchmark in adaptation, lucidity or empathy reveals more than an overall rank: it shows where your approach has an edge.
How it works
An immediate result, with no account.
- 01Find a quiet place where you can focus.
- 02Complete the tasks without outside help.
- 03See your overall score and five dimensions.
- 04Compare your profile with the published ChatGPT benchmark.
- 05Share your strongest advantage and invite someone to do better.
Transparent protocol
The same form, under different conditions.
The GPT benchmark uses the same 29 scored task families and scoring formula as the human assessment. The browser and API versions differ in presentation and timing.
The published result comes from one Codex run with medium reasoning effort, reviewed by M.I. The operator declares the model name; OpenAI does not authenticate it. This specific comparison does not establish general human or AI superiority.
Your turn
ChatGPT has set a score. Discover yours.
Your advantage may not be where you expect it.