How to read the GATE answer key

The official answer key to eighteen of the fifty GATE papers on this site is inside a file you have already downloaded. It is not a separate document and not a coaching solution: it is the organising institute's own key, printed as a table after the last question of the paper. Of the fifty hosted files, eighteen carry that table: CS/IT 2018, 2020, 2022 and 2023; EC 2018, 2020 and 2022; EE 2018, 2020 and 2022; IN 2018, 2020 and 2022; and BT 2018, 2019, 2020, 2022 and 2024. No Data Science and AI paper carries one, and eight files are scans (CS, EC and IN for 2019 and 2021, plus EE 2021 and BT 2021), so nothing is claimed about those either way. Each of the eighteen keys has 65 rows, one per question, which is what the brochure says a paper holds: “Every test paper has 65 questions carrying a total of 100 marks, out of which 10 questions carrying a total of 15 marks will be on General Aptitude (GA).”

Sources: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026; GATE 2027 Information Brochure, p. 23, as on 15 September 2026

What the key gives you, and what it does not

The key gives the accepted answer and the mark value. It gives no working, no method and no explanation, so it tells you that you were wrong and not where. What it does better than any solution book is settle the exam's own conventions — how an answer was accepted, how wide a numerical tolerance was, and which questions were awarded to everybody. The three question types and what a wrong answer costs are set out at /guide/gate-2027-exam-pattern-and-marking and are not repeated here.

Source: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026

Five columns, and a sixth from 2020

The 2018 keys print this header: “Q.No. Type Section Key/Range Marks”. Left to right, that is the question number; the type, one of MCQ, MSQ or NAT; the section, either GA for General Aptitude or the paper's own code such as CS, EC, EE, IN or BT; the accepted answer or the accepted range; and the mark value, 1 or 2. From 2020 onward the keys add a Session column after the question number: the 2020 keys head their columns “Q.No. Session Que.Type Sec. Name Key Marks” and the 2022 keys “Q. No. Session Question Type Subject Name Key/Range Mark”. It records which session's paper each row belongs to, which matters because the brochure normalises multi-session papers. That arithmetic is covered at /guide/how-the-gate-score-is-calculated. One more thing the Section column settles is where a row sits. In the 2018, 2019 and 2020 keys here the numbering restarts at the section boundary — General Aptitude runs 1 to 10 and the subject section runs 1 to 55 — so a paper of that vintage has two rows numbered Q.8 and the section tells them apart; from 2022 the rows run 1 to 65 straight through. The Section column is also the one worth reading twice. It splits your own score into General Aptitude and subject, the split the brochure fixes at 10 questions and 15 marks inside the 65 and 100.

Sources: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026; GATE 2027 Information Brochure, p. 23, as on 15 September 2026; GATE 2027 Information Brochure, p. 27, as on 15 September 2026

MTA: the question everybody got right

Twelve rows across the eighteen keys carry no answer at all. They carry MTA — marks to all. Every candidate who sat that paper received the marks for that question, whatever they entered and whether or not they attempted it. The two 2018 keys that contain one spell it out as “Marks to All”; the 2020, 2022 and 2023 keys use the three-letter code, so searching for the words alone misses ten of the twelve. The key records the outcome, not the reason. Eight of the twelve are MCQ rows, three are NAT and one is an MSQ, and EE 2020 alone carries four — four of that paper's 65 questions measure nothing the candidate did. When you mark an old paper, award yourself an MTA row in full, then read the question anyway. It was a live question in a live exam.

Source: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026

Every marks-to-all row in the eighteen keys

An MSQ can have only one correct option

This is the part of the key that can cost two marks on a question you had right. Of the 67 MSQ rows in the eighteen keys, five accept only one option: CS 2022 Q.52, key A, 2 marks; EC 2022 Q.28, key C, 1 mark; EC 2022 Q.50, key A, 2 marks; EC 2022 Q.51, key A, 2 marks; and BT 2024 Q.53, key C, 2 marks. This is not an oddity in those five keys — it is what the brochure defines. An MSQ “will have choice of four answers, out of which ONE or MORE than ONE choice(s) is/are correct”. The scoring rule on the same page is absolute: “There is NO negative marking for a wrong answer in MSQ. However, there is NO partial mark for choosing partially correct combinations of choices or any single wrong choice.” Put the two together. A candidate who has correctly identified option A, then reasons that a multiple-select question must have more than one answer and ticks B as well, has produced a partially correct combination. It scores zero. The absence of negative marking does not soften that: nothing further is taken away, but the two marks are not earned. The rest of the 67 do take more than one: 37 accept two options and 23 accept three. That is the common case in these keys, which is why the five are worth knowing about — common is not the same as guaranteed.

Sources: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026; GATE 2027 Information Brochure, p. 23, as on 15 September 2026

Tick what you can defend, and nothing more

Sources: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026; GATE 2027 Information Brochure, p. 23, as on 15 September 2026

One key accepts two different answer sets

Only one MSQ row in the eighteen keys accepts two different sets of options: CS 2023 Q.27, a 1-mark MSQ whose key reads “A, C, D or C, D”. Either set was accepted, so a candidate who ticked three options and one who ticked two both scored full marks. It is the only row of its kind among the 67 MSQs here, but the construction appears on other question types too. BT 2019 Q.9 in the BT section and EE 2020 Q.1 in the EE section are MCQ rows keyed “C OR B” and “A OR D”. CS 2020 Q.21 is a NAT row keyed “13.3 to 13.3 OR 13.5 to 13.5”, and EE 2022 Q.63 accepts either 675 to 700 or 1725 to 1740 — two bands that do not touch. If your answer matches either side of the “or”, the key accepts it. One formatting detail: every key here separates the options of an MSQ with commas except BT 2024, which uses semicolons, as in “A;C;D”. The separator carries no meaning.

Source: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026

No key here before 2022 contains an MSQ

Every 2018, 2019 and 2020 key in this set contains zero MSQ rows. The type first appears in the 2022 keys of this set, and not evenly: CS 2022 has 15, EC 2022 has 14, BT 2022 has 8 and IN 2022 has 5, while EE 2022 has none at all, alone among the 2022 papers here. CS 2023 has 15 and BT 2024 has 10. Every MSQ row in all eighteen keys sits in the subject section; not one is in General Aptitude. For practice, an older paper is a full drill in MCQ and NAT technique and no drill at all in the question type that punishes an over-tick. If your revision has run on 2018 and 2020 papers, the first MSQ you meet under time pressure should not be in the exam hall — CS 2022 and 2023, EC 2022, BT 2022 and 2024, and IN 2022 are the papers here that contain them.

Source: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026

Marking an old paper against its own key

1. Mark the paper cold, before opening the key. It tells you that you were wrong, never where, so the diagnosis has to come from your own attempt.
2. Split the result by the Section column. General Aptitude is 10 questions and 15 marks of the 100, and a weak GA half is invisible in a single total.
3. Award yourself any MTA row in full, then read the question anyway. There are twelve of them, four in EE 2020.
4. On an MSQ, compare the two sets, not the two answers. If the key lists one option and you ticked two, the question scores zero.
5. On a NAT row, read the whole band before judging yourself wrong. Of the 406 NAT rows in these keys, 218 — just over half — accept a single exact value, written as “2 to 2”.
6. Read both sides of any “or” in a key. It appears on MCQ, MSQ and NAT rows, and reading only the first half will mark a correct answer wrong.

Sources: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026; GATE 2027 Information Brochure, p. 23, as on 15 September 2026

The key for GATE 2027 itself

None of this is a forecast about 2027. The brochure's post-examination section says that after the examinations conclude the answer keys will be displayed on the official website, and that candidates may submit contests against them during specified dates, subject to payment of a fee. The key you first see for your own paper is a published position, not a closed one. The eighteen keys above are the keys printed in those papers, and reading them costs nothing: the files are free on the previous-year paper pages under /resources. Registration dates, the application fee and eligibility are not covered here — they are on the GATE 2027 application checklist at /resources/gate-2027-application-checklist.

Sources: The 50 hosted paper files, parsed 18 September 2026, as on 18 September 2026; GATE 2027 Information Brochure, p. 27, as on 15 September 2026