Checked 30 September 2026. Counts come from the six official 2024-2026 CS papers and keys and will not change. GATE publishes no subject-wise weightage beyond GA 15, Engineering Mathematics 13 and subject 85, and the 2027 paper can differ.
GATE publishes exactly three weightage figures for Computer Science and Information Technology (CS): General Aptitude 15 marks, Engineering Mathematics 13 marks, and 85 marks for the subject part as a whole, which includes the mathematics. Every other "GATE CS weightage" table you have seen, including the section splits on many coaching sites, is somebody's estimate.
This one is an estimate too, but it shows its working. We took the six official question papers and answer keys for GATE CS 2024, 2025 and 2026 (two sets a year), placed every one of the 330 subject questions into one of the ten sections of the syllabus, and added up the marks from the official key. The classification is ours; the marks are the key's. Where a question could honestly sit in two sections we say which we chose, so you can re-count.
Where other pages get this wrong
No section of GATE CS carried fewer than 5 marks in any of the six sets we counted, so "skip a subject" plans give away real marks.
Commonly published instead: Tables that group four or five subjects as a "Big Five worth 50%+", or print Discrete Mathematics as a separate 11-13% block on top of the official 13 marks for Engineering Mathematics.
Discrete Mathematics is part of Engineering Mathematics in the official syllabus, and the official figure for the whole of Engineering Mathematics is 13 marks. In our count the ten sections averaged between 6.5 and 13.7 marks per set, and the lightest section in any single set, Compiler Design in 2026 CS-1, still carried 5 marks.
Our count of the official 2024-2026 papers and keys. Only the 13 is an official weightage.
What is official, and what is counted
Every GATE CS paper has 65 questions for 100 marks in 3 hours. Questions 1 to 10 are General Aptitude: five 1-mark and five 2-mark questions, 15 marks. Questions 11 to 65 are the subject part: 55 questions for 85 marks, of which the brochure says Engineering Mathematics carries 13. Question types are MCQ, MSQ and NAT. A wrong MCQ costs one third of its marks; MSQ and NAT have no negative marking and no partial credit.
| Part | Questions | Marks | Status |
|---|---|---|---|
| General Aptitude | 10 | 15 | Official (GATE 2027 brochure, Table 8) |
| Engineering Mathematics | not fixed | 13 | Official (GATE 2027 brochure) |
| Rest of the CS syllabus (nine sections) | not fixed | 72 | Official total; the split between sections is not published |
| Whole paper | 65 | 100 | Official |
Marks per section in each of the six official sets
Each cell below is questions and marks for one section in one set. Every column adds up to 55 questions and 85 marks. Sections are ordered by their six-set total.
| Section | 2024 CS-1 | 2024 CS-2 | 2025 CS-1 | 2025 CS-2 | 2026 CS-1 | 2026 CS-2 | Six-set marks | Mean per set |
|---|---|---|---|---|---|---|---|---|
| Engineering Mathematics | 10 Q / 14 m | 9 Q / 13 m | 10 Q / 15 m | 8 Q / 12 m | 10 Q / 15 m | 9 Q / 13 m | 82 | 13.7 |
| Programming & Data Structures | 5 Q / 8 m | 6 Q / 10 m | 6 Q / 9 m | 8 Q / 12 m | 8 Q / 12 m | 6 Q / 10 m | 61 | 10.2 |
| Computer Organization & Architecture | 6 Q / 10 m | 6 Q / 10 m | 5 Q / 8 m | 5 Q / 9 m | 6 Q / 9 m | 5 Q / 9 m | 55 | 9.2 |
| Algorithms | 4 Q / 7 m | 4 Q / 6 m | 6 Q / 9 m | 5 Q / 8 m | 5 Q / 8 m | 7 Q / 10 m | 48 | 8.0 |
| Operating System | 5 Q / 8 m | 5 Q / 8 m | 5 Q / 8 m | 4 Q / 7 m | 5 Q / 8 m | 5 Q / 9 m | 48 | 8.0 |
| Computer Networks | 6 Q / 9 m | 6 Q / 9 m | 4 Q / 6 m | 5 Q / 6 m | 5 Q / 8 m | 6 Q / 9 m | 47 | 7.8 |
| Databases | 6 Q / 8 m | 6 Q / 8 m | 5 Q / 8 m | 5 Q / 9 m | 4 Q / 6 m | 4 Q / 6 m | 45 | 7.5 |
| Digital Logic | 4 Q / 6 m | 4 Q / 6 m | 5 Q / 8 m | 6 Q / 9 m | 5 Q / 8 m | 5 Q / 7 m | 44 | 7.3 |
| Theory of Computation | 4 Q / 7 m | 4 Q / 7 m | 5 Q / 8 m | 5 Q / 7 m | 4 Q / 6 m | 4 Q / 6 m | 41 | 6.8 |
| Compiler Design | 5 Q / 8 m | 5 Q / 8 m | 4 Q / 6 m | 4 Q / 6 m | 3 Q / 5 m | 4 Q / 6 m | 39 | 6.5 |
| 2024 S1 | 2024 S2 | 2025 S1 | 2025 S2 | 2026 S1 | 2026 S2 | |
|---|---|---|---|---|---|---|
| Engineering Mathematics | 14 | 13 | 15 | 12 | 15 | 13 |
| Programming & Data Structures | 8 | 10 | 9 | 12 | 12 | 10 |
| Computer Organization & Architecture | 10 | 10 | 8 | 9 | 9 | 9 |
| Algorithms | 7 | 6 | 9 | 8 | 8 | 10 |
| Operating System | 8 | 8 | 8 | 7 | 8 | 9 |
| Computer Networks | 9 | 9 | 6 | 6 | 8 | 9 |
| Databases | 8 | 8 | 8 | 9 | 6 | 6 |
| Digital Logic | 6 | 6 | 8 | 9 | 8 | 7 |
| Theory of Computation | 7 | 7 | 8 | 7 | 6 | 6 |
| Compiler Design | 8 | 8 | 6 | 6 | 5 | 6 |
Shading: 12 or more marks darkest, 9-11 dark, 7-8 light, 6 or fewer palest.
Our classification of the official papers; marks from the official keys.
Two readings matter more than the ranking. First, the spread is narrow: after Engineering Mathematics, the other nine sections sit between 6.5 and 10.2 marks per set on average, so the difference between the third and the eighth section is about two marks a paper. Second, the sets inside one year differ as much as the years do. In 2026, Algorithms carried 8 marks in the forenoon set and 10 in the afternoon set; Programming and Data Structures went 12 and 10. You write one set, and you do not choose which.
What the six sets say
- Engineering Mathematics is the largest block at 12 to 15 marks per set in our count, a mean of 13.7 against the official 13. The gap is graph-theory and recurrence questions that could equally be counted under Algorithms.
- Programming and Data Structures is next at 8 to 12 marks, most of it C code tracing, recursion, trees and heaps.
- Computer Organization carried 8 to 10 marks every set, and it is the most numerical section: 15 of its 33 questions were NAT.
- Compiler Design is the lightest and drifting down: 8, 8, 6, 6, 5 and 6 marks across the six sets.
- Databases fell to 6 marks in both 2026 sets, after 8 to 9 in the four earlier sets.
Six sets over three years is a small sample. These are tendencies, not promises about 2027.
Question types by section
Negative marking only touches MCQs, so how a section is usually asked changes how you should attempt it. This table adds up all six sets.
| Section | MCQ (Q / marks) | MSQ (Q / marks) | NAT (Q / marks) | Marks with no negative marking |
|---|---|---|---|---|
| Engineering Mathematics | 20 / 23 | 16 / 27 | 20 / 32 | 59 of 82 (72%) |
| Programming & Data Structures | 15 / 23 | 9 / 14 | 15 / 24 | 38 of 61 (62%) |
| Computer Organization & Architecture | 12 / 17 | 6 / 8 | 15 / 30 | 38 of 55 (69%) |
| Algorithms | 15 / 23 | 7 / 11 | 9 / 14 | 25 of 48 (52%) |
| Operating System | 6 / 9 | 10 / 15 | 13 / 24 | 39 of 48 (81%) |
| Computer Networks | 16 / 23 | 8 / 9 | 8 / 15 | 24 of 47 (51%) |
| Databases | 12 / 17 | 13 / 19 | 5 / 9 | 28 of 45 (62%) |
| Digital Logic | 7 / 11 | 15 / 22 | 7 / 11 | 33 of 44 (75%) |
| Theory of Computation | 8 / 11 | 14 / 22 | 4 / 8 | 30 of 41 (73%) |
| Compiler Design | 15 / 24 | 6 / 9 | 4 / 6 | 15 of 39 (38%) |
Counted from the official answer keys. Only 27 to 36 of the 85 subject marks in any set carried negative marking.
Operating System, Computer Organization and Engineering Mathematics lean numerical; Digital Logic and Theory of Computation lean towards MSQ, where every correct option must be marked and nothing else. Compiler Design was the most MCQ-heavy (24 of its 39 marks), followed by Networks and Algorithms at about half, so they are where a careless guess costs most. None of this is a rule of the exam; it is how the last six sets were written.
Where the 2027 syllabus makes the old numbers misleading
The 2027 CS syllabus is not the 2026 one. Our comparison of the two official PDFs found seven sections unchanged and three rewritten: Digital Logic, Computer Organization and Architecture, and Computer Networks. That matters here because past weightage is only a guide to the topics still examinable.
| Section | Six-set mean | What changed for 2027 (our reading) | Effect on the count |
|---|---|---|---|
| Computer Networks | 7.8 marks | OSI and TCP/IP stacks, framing, bridging, ARP, DHCP, ICMP, UDP, SMTP, FTP and Email are no longer listed | At least four 1-mark questions in the six sets rested mainly on topics no longer listed, so the past mean overstates what is now in scope |
| Computer Organization and Architecture | 9.2 marks | Hardwired and microprogrammed control, memory interfacing and cache mapping are named; data-path, main memory and secondary storage are not | Disk questions (three in six sets) were filed under COA; secondary storage is no longer named in COA |
| Digital Logic | 7.3 marks | K-map, the tabular method and circuit design are named explicitly | Scope is, if anything, more explicit; no past question falls outside it |
The full word-by-word comparison, with the past questions that now fall outside the syllabus, is in GATE CS 2027 syllabus changes explained.
How we classified, and every judgement call
Each question went into the one section whose official syllabus wording best describes what it tests. Where two sections fit, we followed a fixed rule: a question goes where the concept being tested is taught, not where its setting comes from. So a recurrence solved to a Theta bound counts as Algorithms, a spanning-tree count of a complete graph counts as Engineering Mathematics (graph theory), and a disk-transfer calculation counts as Computer Organization. The 24 items below are the ones we flagged; moving every one of them would shift no section by more than a few marks over six sets.
| Set | Questions, our section (other possible section) |
|---|---|
| 2024 CS-1 | Q33 operator precedence: Compiler Design (Programming); Q34 spanning trees of K4: Maths (Algorithms); Q43 heap index count: Programming (Algorithms); Q54 hard-disk geometry: COA (OS); Q55 multicore speed-up: COA (flagged, no clear second section) |
| 2024 CS-2 | Q15 linear recurrence: Algorithms (Maths); Q42 C code, longest subarray: Programming (Algorithms); Q51 spanning-tree weight parity: Algorithms (Maths); Q53 disk transfer time: COA (OS) |
| 2025 CS-1 | Q43 BFS tree height vs diameter: Algorithms (Maths); Q56 bit-flip probability on a channel: Maths (Networks); Q59 state minimisation: Digital Logic (Theory of Computation) |
| 2025 CS-2 | Q32 Booth multiplication: Digital Logic (COA); Q38 meld complexity: Programming (Algorithms); Q51 if-then-else grammar ambiguity: Compiler Design (Theory of Computation) |
| 2026 CS-1 | Q25 grammar ambiguity: Theory of Computation (Compiler Design); Q47 vertex cover: Maths (Algorithms); Q54 TLB, page table and cache: OS (COA); Q55 2-colourability: Maths (Algorithms); Q59 disk transfer time: COA (OS); Q61 recursive functions: Programming (Algorithms) |
| 2026 CS-2 | Q17 what lives on the heap in C: Compiler Design (Programming); Q29 which grammars are ambiguous: Theory of Computation (Compiler Design); Q36 spanning trees of Kn: Maths (Algorithms) |
Our count keeps these rules consistent across all six sets, so the trends are comparable even where a single placement is arguable. If you prefer the other section for any item, subtract its marks from our choice and add them to yours; the question numbers are the official ones, so you can check each item in the papers.
Turning the count into a study order
| Order | Section | Why |
|---|---|---|
| 1 | Engineering Mathematics | Largest block, official 13 marks; discrete maths also powers Algorithms and TOC |
| 2 | Programming & Data Structures | 8-12 marks a set, heavy on code tracing that practice fixes quickly |
| 3 | Computer Organization & Architecture | 8-10 marks every set, mostly numerical; the 2027 wording adds control-unit design and cache mapping |
| 4 | Operating System | 7-9 marks, the most NAT-heavy section after COA |
| 5 | Algorithms | 6-10 marks, and shares ground with maths and data structures |
| 6 | Digital Logic | 6-9 marks; K-map and the tabular method are now named |
| 7 | Computer Networks | 6-9 marks in the past, but its 2027 syllabus is smaller |
| 8 | Databases | 6-9 marks; SQL, normal forms and indexing recur |
| 9 | Theory of Computation | 6-8 marks, MSQ-heavy, so precision matters |
| 10 | Compiler Design | Lightest, 5-8 marks, and trending down |
Order means the sequence in which to make each section solid, not permission to drop the bottom of the list. In our count the gap between the fourth and the ninth section is under two marks per set, and small sections are often the cheapest marks because their syllabus is short. The last three sections in this order together still carried 17 to 23 marks in every set (our count).
Weightage says where marks sit; it does not say how many you need. For that, see GATE CS qualifying marks and score vs rank, and for the papers themselves, GATE CS previous year papers with answer keys.
Practise it, don't just read it
Weightage tells you where the marks are. A topic quiz tells you whether they are yours yet.
The ProSyllabus GATE CS library currently has 29 topic quizzes of ten questions each, covering programming and data structures, algorithms, operating systems, databases, computer networks, compiler design and theory of computation, each with a worked explanation. There are no quizzes yet for Engineering Mathematics or Computer Organization, and Digital Logic has only number systems, so use the official papers for those. One networks quiz is built on the OSI and TCP/IP stacks, which the 2027 syllabus no longer lists.
Which subject has the highest weightage in GATE CS?
Engineering Mathematics, which the GATE 2027 brochure fixes at 13 of the 85 subject marks. In our count of the six official 2024-2026 sets it averaged 13.7 marks, followed by Programming and Data Structures at 10.2 and Computer Organization at 9.2.
Does GATE publish subject-wise weightage for CS?
No. The only official figures are General Aptitude 15 marks, Engineering Mathematics 13 marks and 85 marks for the subject part. Any section-by-section table, including ours, is a count or estimate made from past papers.
Is Compiler Design still worth studying for GATE CS?
Yes. It is the lightest section in our count, but it carried 5 to 8 marks in every one of the six sets, and its syllabus is unchanged for 2027.
Did Computer Networks weightage change for 2027?
The syllabus did. The 2027 Computer Networks section no longer lists the OSI and TCP/IP stacks, ARP, DHCP, ICMP, UDP, SMTP, FTP or Email. Past networks marks therefore include some topics that are no longer in scope, so treat the past average of 7.8 marks as an upper guide.
How many questions are there in GATE CS?
Sixty-five questions for 100 marks in 3 hours: 10 General Aptitude questions for 15 marks and 55 subject questions for 85 marks, which include Engineering Mathematics.
More in this series
- GATE CS previous year papers with official answer keys
- GATE CS qualifying marks, score vs rank, with calculator
- GATE CS 2027 syllabus changes, section by section
- PSU recruitment through GATE CS 2027, from the notices
- GATE DA vs GATE CS: which paper to write
- GATE CS rank predictor
- GATE CS expected cut-off for top IITs
- GATE CS notice board: official updates
- GATE CS exam analysis
- GATE Computer Science complete guide
Sources
- official — GATE 2027 Information Brochure, IIT Madras, revised 27 Sep 2026
- official — GATE 2027 question paper pattern
- official — GATE 2027 CS syllabus (IIT Madras)
- official — GATE 2026 CS syllabus (IIT Guwahati)
- official — GATE 2026 CS-1 question paper
- official — GATE 2026 CS-1 Answer Key
- official — GATE 2026 CS-2 question paper
- official — GATE 2026 CS-2 Answer Key
- official — GATE 2025 CS-1 question paper
- official — GATE 2025 CS-1 Answer Key
- official — GATE 2025 CS-2 question paper
- official — GATE 2025 CS-2 Answer Key
- official — GATE 2024 CS-1 question paper
- official — GATE 2024 CS-1 Final Answer Key
- official — GATE 2024 CS-2 question paper
- official — GATE 2024 CS-2 Final Answer Key
official = a document published by the conducting body. reported = a news or coaching site we could not check against an original. Where sources disagree this page says so rather than picking one.
