Every number we advertise, and how it was counted
Question banks in this market advertise figures nobody can check. Ours are read straight from the database that serves you, and printed here with the date they were read. If a number on an ad doesn’t match a number on this page, the ad is wrong — tell us and we’ll fix it.
Last audited 13 August 2026.
From 500,000+ raw questions to 110,589 you can be served
- 500,000+sourced
Raw questions collected from past papers, published banks and real exam material — before any cleaning.
- 141,771in the bank
What survived cleaning and de-duplication. Exact duplicates, empty stems and broken options are removed at import, not hidden.
- 31,182pulled
Questions a second, independent model disputed, that had no defensible key at all, or that failed review for another reason — corrupted numbers, or a fact that has since gone stale. Excluded from every draw.
- 110,589live
Everything you can actually be served. 104,993 of these have been machine-cross-checked and agreed; the rest are queued for the next wave and are never used in an MPT mock.
What “verified” actually means here
A second model, not the one that wrote anything
Each answer key is re-derived independently and compared with the key the source shipped. Agreement marks the question cross-checked. Disagreement does not silently “fix” it.
Disputes are pulled, never auto-corrected
31,182 questions are quarantined — removed from every draw. Most are there because a second model disputed the key; the rest had no usable key, corrupted figures, or a fact that has gone stale. Automatically rewriting a key is how a bank ends up confidently wrong, so a disputed question waits for a human instead of being guessed at. You are never shown one.
The MPT mock draws only from cross-checked questions
The full-length MPT and GSA mocks are restricted to the 104,993 questions that passed cross-checking. A question that is merely “not yet disputed” is not good enough for a paper you are using to judge whether you are ready.
And you are the last check
Every question in the app has a report button. Reports go to a queue a human reads, and a question that turns out to be wrong is quarantined like any other — pulled from every draw across all 1,562 topics, not quietly edited. Machines find the disagreements at scale; people still decide.
What we don’t claim
A page about honesty that only lists strengths isn’t one. These are the limits of what this product does.
We don’t promise you a score
No product can. Your readiness signal here is an honest weighted average of how you have actually performed, topic by topic — a study aid, not a prediction, and never dressed up as psychometrics.
We don’t claim the whole bank is MPT material
110,589 is the size of the entire live bank across every subject we cover. The MPT-syllabus slice is much smaller, and the app only ever draws MPT questions from that slice.
We don’t simulate the CSS written papers
The engine drills MCQs. It cannot evaluate an essay, a précis or a subjective answer, so it does not pretend to — for the written exam you still need a teacher.
Current Affairs is thin, and we say so in the product
Current Affairs depth is our weakest area. Rather than fill it with model-generated questions — which is exactly how wrong answer keys are born — thin topics are flagged in the coverage map until real, source-grounded questions replace them.
Found something that doesn’t add up? Tell us. Corrections to this page are published with the date they were made.
Take a full-length MPT mock — free
Scored exactly like the real screening test: 200 questions, 200 minutes, no negative marking. See your real score and where you stand.
Free account · No card required · Academies charge Rs 20,000–30,000/month