GATE Computer Science (CS): Paper, Cut-offs, Score & Syllabus

GATE Computer Science is not just the biggest paper in the exam — it is bigger than the next two put together. 259,922 candidates registered for it in GATE 2026, against 115,448 for Electronics and 97,486 for Civil. That is one in four of everyone who registered for GATE at all, and it has grown 70% in two years.

Size changes the mechanics. CS is the only paper of the five covered in this series that is run in two separate sessions, which means your raw marks are not what your score is computed from — normalised marks are. And unlike Mathematics, Physics, Chemistry and Statistics, the CS paper keeps an Engineering Mathematics section worth 13 marks. With General Aptitude on top, 28 of your 100 marks are not computer science.

This article is about the Computer Science paper specifically: what is in it, what the qualifying mark and the GATE score actually measure, what the syllabus contains, and what IIT Madras has changed for GATE 2027.

If you have not sat GATE before, the general guide to the exam covers the rules common to all thirty papers — the marking scheme, eligibility and what a score is used for. This page assumes them.

Which year is this data from? The exam pattern, marking scheme, qualifying-mark formula, score formula, eligibility and fee are from the GATE 2026 Information Brochure published by IIT Guwahati. Candidate numbers and qualifying marks are from the official cut-off pages for GATE 2024, 2025 and 2026 — published by IISc Bengaluru, IIT Roorkee and IIT Guwahati respectively. The question-type breakdown is counted from the official GATE 2025 answer key.

GATE 2027 is organised by IIT Madras, and the syllabi have been revised. What changed for this paper is set out in section 5 below. This page is updated after each cycle rather than replaced, so the older figures stay here beside the new ones.

1. One exam, three destinations

Before the paper itself, the thing GATE is for. One score does three different jobs, and they do not treat it the same way.

What it is used for	How the score is treated
M.Tech, M.E., M.S. and M.Sc. at IITs, IISc, NITs and other centrally funded institutes	A three-year-old score is accepted. Many programmes carry a monthly stipend.
PhD admission at the same institutes	A three-year-old score is accepted, often directly after a bachelor’s degree.
Recruitment by public sector companies	Usually the current year’s score only.

That last row costs people a year. Your scorecard is valid for three years and institutes honour it; most PSU recruitment notices ask for the score from that year’s exam.

2. What is actually in the paper

Every GATE paper is the same shape: 65 questions, 100 marks, three hours, on a computer, in English. General Aptitude is 15 marks and is identical for all 30 papers; the remaining 85 are yours.

What differs between papers, and between years, is the mix of question types — and that mix decides how much of the paper you can attempt without risk. Here is GATE 2025 Computer Science Session 1, counted from the official answer key.

Section	Questions	Marks	Negative marking
General Aptitude — MCQ only	10	15	Yes, on every question
Subject — MCQ	20	30	Yes
Subject — MSQ	13	18	None
Subject — NAT	22	37	None
Total	65	100

Question type	What it is	Wrong answer costs
MCQ	Four options, exactly one correct	−1/3 on a 1-mark question, −2/3 on a 2-mark question
MSQ	One or more options correct	Nothing — but there is no partial credit either
NAT	Type the number; no options given	Nothing

55 of the 85 subject marks carried no penalty at all. In an MSQ or a NAT, a blank and a wrong answer score exactly the same — zero — so leaving one untouched costs precisely as much as an attempt that turns out wrong, and gains nothing.

MSQ carries its own trap in the other direction. With no partial credit, knowing two of four options are correct is worth nothing unless you can also rule on the other two. And an MSQ is not obliged to have more than one correct option: in GATE 2025 Computer Science Session 1, 3 of the 13 multiple-select questions had exactly one correct answer. Stopping after the first correct tick is how a well-prepared candidate scores zero.

One more line worth carrying: all 15 marks of General Aptitude are negatively marked, because the whole section is MCQ. It is the only part of the paper that is guaranteed to punish a guess.

3. What the qualifying mark actually is

Most people read the GATE cut-off as a target set by an examiner. It is not. It is a formula, and the brochure prints it:

Qualifying mark for the General category = max(25, min(40, μ + σ))

where μ is the mean and σ the standard deviation of the marks of everyone who appeared in that paper.

Two consequences follow immediately, and they change how you should read the number.

The cut-off is a description of the cohort, not of the paper’s difficulty. It is literally one standard deviation above the average mark. If it comes out low, the average candidate scored low; if it comes out high, they scored well.

And it is clamped between 25 and 40. A paper whose μ + σ falls below 25 publishes 25 anyway, and one whose μ + σ exceeds 40 publishes 40. When you see either of those exact numbers, the formula has been overruled by the floor or the ceiling, and the published figure tells you nothing about the cohort beyond “below 25” or “above 40”. The ceiling is recent: in GATE 2024 six papers qualified above 40, with one at 49.0. Since GATE 2025 no paper has exceeded 40.

Year	Registered	Appeared	Turnout	Cut-off (GEN)	OBC-NCL / EWS	SC / ST / PwD
GATE 2024	152,652	123,967	81.2%	27.6	24.8	18.4
GATE 2025	207,851	170,825	82.2%	29.2	26.2	19.4
GATE 2026	259,922	211,020	81.2%	30	27	20

The other two columns are not set separately. The brochure fixes them by ratio: OBC-NCL and EWS at 9/10 of the General mark, and SC, ST and PwD at 2/3 of it. Published figures are rounded to one decimal, so the arithmetic can look 0.1 out; the rule itself is exact.

Two sessions, and why your raw marks are not your marks. Because CS is too large for one sitting, it runs in two sessions with two different papers. The brochure sets out the normalisation: raw marks are rescaled using the average of the top 0.1% in each session and in the paper overall, so that a harder session is not a disadvantage. Your GATE score is computed from normalised marks, not from what you scored.

The practical consequence is that comparing your raw marks against a friend in the other session, or against last year’s paper, tells you nothing useful. The GATE 2025 sessions were not even built the same way: Session 1 had 20 MCQ, 13 MSQ and 22 NAT; Session 2 had 22, 18 and 15.

4. What a GATE score actually means

A GATE score is not your marks, and it is not a percentile. It is a linear rescaling with two fixed anchors, and once you know them the whole thing is readable:

• A score of 350 is exactly the qualifying mark. If you scraped through, you scored 350.
• A score of 900 is the average of the top 0.1% of candidates in your paper (or the top 10, whichever is larger).

Everything between is a straight line. The formula in the brochure is GATE Score = 350 + 550 × (your marks − Mq) ÷ (top-0.1% average − Mq), where Mq is the General-category qualifying mark.

This is why a score is comparable across papers in a way that marks are not. Two candidates on 700 in different subjects sit at the same point between their own paper’s bar and its top. It is also why the score depends on how well the best candidates did that year: if the top 0.1% average is lower, the same marks are worth more.

The scorecard is valid for three years. Institutes honour that; most public sector companies do not — their recruitment notices usually ask for the current year’s score. If a PSU job is the goal, the year you sit is part of the plan.

5. The syllabus, and what changes in GATE 2027

Ten sections — and the first one is not computer science. Engineering Mathematics carries 13 of the 85 subject marks: discrete mathematics, combinatorics, linear algebra, calculus, probability and statistics. Add the 15 marks of General Aptitude and 28 of your 100 marks come from outside the discipline.

Candidates routinely under-prepare both, on the reasoning that the real paper is the other nine sections. In an exam where ranks separate on fractions of a mark, 28 marks is not a rounding error, and it is the most predictable part of the paper — discrete mathematics and probability do not change from year to year the way a systems question does.

What IIT Madras changed for GATE 2027

The GATE 2027 Computer Science syllabus keeps all ten sections in the same order. The revision is a tightening rather than a restructuring — of the five papers in this series, CS is the one that changed least. Even so, the syllabus is revised between cycles as a matter of routine, and topics do get added and removed.

Read your paper’s official syllabus for the year you are sitting before you buy a single book. Working from a coaching module that predates the current syllabus is a common and entirely avoidable way to study topics the exam has dropped.

6. Who may sit it, and what it costs

Rule	What it says
Qualifying degree	Currently in the third year or higher of any undergraduate degree, or already holding one.
Degree subject	Need not match the paper. You choose the paper; the admitting institute sets its own subject requirements.
Age limit	None.
Attempts	No limit.
Two papers	You may sit up to two papers in one cycle, from the published list of allowed combinations.
Application fee (per paper)	₹1,000 for female, SC, ST and PwD candidates · ₹2,000 for everyone else. Both rise by ₹500 in the extended period.

The third-year rule is worth acting on rather than noting. You do not need the degree in hand, and a good score obtained in your third year is valid for three years — which covers the entire window in which you might want to use it.

7. What this paper is asking for

1. Attempt every MSQ and every NAT. Fifty-five of the 85 subject marks carried no penalty in GATE 2025 Session 1 — and 37 of those were numerical-answer questions.
2. Check every option on an MSQ. No partial credit — and three of the thirteen MSQs in Session 1 had exactly one correct option.
3. Prepare the 28 marks that are not computer science. Engineering Mathematics (13) and General Aptitude (15) are the most predictable marks on the paper and the most commonly neglected.
4. Do not chase the other session’s paper. Normalisation exists precisely so that a harder session is not a disadvantage. Comparing raw marks across sessions tells you nothing.
5. Breadth beats depth here. Ten sections over 55 subject questions means roughly five or six questions per section. Depth in algorithms cannot cover a hole in operating systems.
6. Aim at a score, not at the bar. With 211,020 candidates in the room, 350 clears the qualifying mark and nothing else.

The syllabus is public and 211,020 people have the same one. What separates results is the sections other people skipped.