OpenGPTEvidence-based model decisions
Help
PRIVATE MANUAL A/B TEST · NO API

Compare two models on the same practical tests

Run the same prompt in both models, paste the two answers here, and record a structured A/B decision for each test.

✓ No API callsOpenGPT uploads nothing; candidate services still receive prompts you submit to them.

Choose two models

Select the exact versions you want to compare.

Choose one real-work template

Each template contains three short, repeatable tests.

Saving draft on this device…
Progress0 / 3 completed
0 / 3
1 / 3

Return eligibility

Run this test

  1. 1Copy the prompt
  2. 2Run it in Model A
  3. 3Run it in Model B
  4. 4Paste both answers
  5. 5Choose the better answer
1

Test prompt

Policy: Unopened products may be returned within 30 days. Opened software is non-refundable. Mina bought an unopened keyboard 18 days ago. Write a concise, helpful reply that states eligibility and the next step.

Select two different versions first.

2Answer AModel A
model service

Select this exact model: —
Paste the copied prompt, then run it. Paste the complete answer below.

3Answer BModel B
model service

Select this exact model: —
Paste the copied prompt, then run it. Paste the complete answer below.

Which answer is better?
Paste both answers before choosing a verdict.