CBSE Class 10 · Artificial Intelligence, code 417 · Previous year papers, 2025 and 2026

Class 10 AI previous year question papers, read closely and answered in fresh practice.

This class 10 AI previous year question paper guide tells you what the CBSE 417 board papers of 2025 and 2026 actually asked, section by section, in our own words. We do not reproduce the papers. Instead we show which shapes come back every year, map them to the 2026-27 units, and give original questions in those shapes with the working an examiner wants to see. In our Monday and Wednesday batch these shapes are rehearsed until they feel routine.

Book a free demo class Book a priority demo Ask on WhatsApp

Live online · 2025 and 2026 papers described · repeats mapped by unit · original practice solved · full timed papers in term two · first class free

Start here

Courses behind this past-paper practice

The 417 course whose Class 10 batch works through these question shapes, the Python course for the practical half of the subject, and the Orange course for the no-code tools the newer papers name in their options.

CBSE Computational Thinking and AI for Teens course thumbnail

The batch / 417 Class 10

CBSE AI for Class 9 to 12

The full 417 course: each unit taught, then tested in the exact shapes the board has used, with answers marked against the examiner value points.

Open the course →
Python for Teens course thumbnail

Practical half / Python

Python for Teens

The NumPy, Matplotlib and CSV programs that the practical file and the lab exam need, which no theory paper asks you to write.

Open the course →
Visual Machine Learning with Orange for Teens course thumbnail

No-code tools / Orange

Visual ML with Orange

Hands-on with the no-code tool behind the penguin, coral and sentiment case studies, so a scenario question about them reads as familiar.

Open the course →

Ans. The short version

For a class 10 AI previous year question paper, start with the CBSE 417 main papers of 2025 and 2026. Both had 21 questions, of which a student answers 15, in two hours for 50 marks: 24 objective marks in Section A and 26 written marks in Section B. Every paper we read had a four-mark confusion matrix case, a four-mark Bag of Words table, a stemming question, a chatbot type and assertion-reason items. The 2025 paper was set on the older syllabus. Modern Age Coders rehearses these shapes live on Monday and Wednesday at 9:00 PM IST, or one to one.

Q1. What did each year ask?

What did the CBSE Class 10 AI board papers ask in 2025 and 2026?

Both papers used the same frame: 21 questions, 15 to answer, two hours, 50 marks. What changed between them was the syllabus underneath. Here is each paper described in our words, from one set per year that we read in full.

PaperSyllabus it followedWhat stood outHow we checked it
2026 board exam (main, 27 February 2026)The current units: project cycle and ethics, modeling, evaluating models, computer vision, NLPA cyber-security case asking for a confusion matrix, one cell and precision; a Bag of Words table from two short documents; AI, ML and DL compared with a Venn diagram; two scenarios to match to an AI domain; stemming against lemmatisation on a plural noun; sector-based against value-based ethicsMarking scheme for set 4 of paper code 104, read line by line
2025-26 sample paperThe current unitsA biased admissions algorithm judged with bioethics principles; unsupervised learning with clustering and association; four scenarios to name by model type; a matrix built from a story, then accuracy and wrong predictions; a Bag of Words dictionary and vector with two short theory partsCBSE's own sample paper marking scheme
2025 board exam (main)The older units, which still had Data Science and Advance Python in theoryA diagnostic-test case asking for true negatives, precision, recall and F1; a Bag of Words table; a neural network diagram to explain; the problem statement template; AI bias and AI access with examples; items on SQL, CSV files, IoT and data science in airlinesCBSE's own marking scheme, set 4 of paper code 104
2024 and earlierOlder units again, with a different section design in some yearsNot described hereNot verified by us, so we make no claims about them

What the frame looked like in both years

  • Section A, 24 marks. Five questions of six one-mark parts each: any 4 of 6 from employability skills, then any 5 of 6 in each of the four AI questions.
  • Section B, 26 marks. Any 3 of 5 employability two-markers, any 4 of 6 AI two-markers, any 3 of 5 AI four-markers.
  • Word limits printed on the paper: 20 to 30 words for two marks, 50 to 80 for four.
  • No negative marking in Section A in either year.

Assertion-reason is already here

Each of the three papers above carried assertion-reason items inside Section A, alongside two-statement items where you judge Statement 1 and Statement 2 separately. So these formats have appeared in 417 papers, not just in IT 402. CBSE has not said how many a 2027 paper will carry, so treat them as one of several Section A shapes, not the main one. A full practice set is on assertion and reason questions for Class 10 AI.

The blueprint and the 2027 calendar are on the exam date and paper pattern page.

Q2. What comes back every year?

Most repeated questions in class 10 AI, tagged by the 2026-27 unit

A repeat here means the same skill was tested in the same shape, never the same wording. Each row was present in all three papers we read unless the cell says otherwise.

Repeating shapeUnit now2025 board2025-26 sample2026 board
Confusion matrix case: build or read the matrix, then a metric3 Evaluating Models4 marks4 marks4 marks
Bag of Words: dictionary and document vectors from short sentences6 NLP4 marks4 marks4 marks
Stemming against lemmatisation on one word6 NLP1 mark1 mark2 marks
Script bot or smart bot6 NLP1 mark1 mark1 mark
False positive or false negative named from a situation3 Evaluating Models1 and 2 marks1 mark1 mark
Overfitting, or why test data must be unseen3 Evaluating Models1 mark1 mark1 and 2 marks
Neural network layers, or CNN and ANN2 Modeling4 marks2 marks4 marks
Learning types: supervised, unsupervised, clustering, association2 Modeling1 mark4 marks1 and 2 marks
Computer vision basics: tasks, pixel values, resolution, RGB5 Computer Vision1 and 2 marks1 and 2 marks1 and 2 marks
Ethics: bias, bioethics principles, framework types1 Project cycle and ethics4 marks4 marks1 and 2 marks
Match a scenario to an AI domain or model1 and 21 mark4 marks4 marks
Part A: communication, self-motivation, entrepreneur tasks, sustainabilityPart A1 and 2 marks1 and 2 marks1 and 2 marks

The two safe four-markers

The matrix case and the Bag of Words table appeared as four-mark questions in every paper we read. Together they are 8 of the 12 long-answer marks a student needs, so they are worth practising until they take under eight minutes each.

The rising unit

Ethics moved from a single essay on bias to a mix: a value-based or sector-based item, an assertion on bioethics and a long case. Unit 1 now rewards knowing the four bioethics principles by name and applying them.

The thin unit

Computer Vision has only four theory marks, and it was asked as one-mark facts and a two-mark explanation. Pixel arithmetic and RGB storage cover nearly all of it.

Fresh numbers for the matrix case are on confusion matrix numericals for Class 10 AI, and longer corpora with TFIDF on Bag of Words and TFIDF, worked.

Q3. Section A, in the board's shapes

CBSE class 10 AI question paper practice with solutions: Section A shapes

Ten original one-mark items, one for each format the recent papers used: plain MCQ, a quick calculation, assertion-reason, two statements and one-word answers. Try each before opening the answer.

Items 1 to 5

  1. A model is tested on 500 samples: TP 150, TN 280, FP 40, FN 30. How many predictions were correct?
    (a) 180
    (b) 430
    (c) 320
    (d) 70
    Answer (b) Correct means TP + TN = 150 + 280. Option (d) is the number of wrong predictions, FP + FN.
  2. Out of 60 emails a filter marked as spam, 45 really were spam. Its precision is
    (a) 0.45
    (b) 0.60
    (c) 0.75
    (d) 1.33
    Answer (c) Precision = TP / all predicted positive = 45 / 60. Recall cannot be found: we are not told how many spam emails were missed.
  3. Assertion (A): A rule-based chatbot can answer only the questions its makers wrote replies for. Reason (R): A rule-based system does not learn from new data on its own.
    (a) Both true, R explains A
    (b) Both true, R does not explain A
    (c) A true, R false
    (d) A false, R true
    Answer (a) It is limited because it cannot learn, so the reason is the cause of the claim.
  4. Statement 1: Lemmatisation always returns a real dictionary word. Statement 2: Stemming is faster because it only cuts off word endings.
    (a) Both correct
    (b) Both incorrect
    (c) Only 1 correct
    (d) Only 2 correct
    Answer (a) Judge each statement alone. They do not have to explain each other in this format.
  5. A screen is 1280 x 720 pixels. How many pixels does it show?
    (a) 2,000
    (b) 92,160
    (c) 921,600
    (d) 2,764,800
    Answer (c) 1280 x 720 = 921,600. Option (d) multiplies by 3 again, which counts RGB values, not pixels.

Items 6 to 10

  1. A model groups news articles into topics with no labels given. Name the learning type and the subcategory.
    Answer Unsupervised learning, clustering. Name both: the question asks for two things.
  2. A bank's model lets a fraudulent payment through as genuine. Which outcome is this?
    Answer False negative: reality yes (fraud), prediction no.
  3. Which value of the four bioethics principles is broken when an AI shares a patient's records without asking them?
    Answer Respect for autonomy: the patient's right to decide about their own data and care.
  4. Assertion (A): Accuracy alone can hide a weak model. Reason (R): When one class is rare, a model that always predicts the common class still scores high accuracy.
    Answer Both true, and R explains A. This is why the papers also ask for precision, recall or F1.
  5. Inputs 1, 0 and 1 reach a perceptron with weights 0.5, 0.2 and 0.4 and a bias of -0.6. What is the weighted sum plus bias?
    Answer 0.5 + 0 + 0.4 - 0.6 = 0.3. Write the three products first, then add the bias, so a slip is easy to find.

Thirty-six more objective items, grouped by unit, are on Class 10 AI MCQ with answers.

Q4. The four-mark shapes, solved

CBSE class 10 AI question paper 2026 with solutions, the way the long answers repeat

Two original four-mark questions in the two shapes that every paper we read included, answered the way the value points split the marks: half a mark for each formula, half for substitution, the rest for the result or the table.

Case: a crop leaf scanner4 marks
A phone app checks 400 leaf photos for
blight. It says "blight" for 150 photos;
120 of those really have it. Of the 250 it
calls healthy, 40 actually have blight.

(a) Draw the confusion matrix.      1
(b) Find precision and recall.      2
(c) Find the F1 score.              1

              Reality: Yes  Reality: No
Predicted Yes   TP = 120     FP = 30
Predicted No    FN = 40      TN = 210

Precision = 120 / (120 + 30) = 0.80
Recall    = 120 / (120 + 40) = 0.75
F1 = 2 x 0.80 x 0.75 / (0.80 + 0.75)
   = 1.20 / 1.55 = 0.77 (approx.)

Build TN last: 250 called healthy minus the 40 missed gives 210. The check is that all four cells add to 400. An F1 of 0.775 means the plain average was used, which loses the mark.

Bag of Words, all four steps4 marks
Document 1: Rain helps crops grow.
Document 2: Crops need rain and sun.

Step 1  Normalise: lower case, drop the
        full stops, split into tokens
  D1: [rain, helps, crops, grow]
  D2: [crops, need, rain, and, sun]

Step 2  Dictionary of unique words
  rain helps crops grow need and sun

Step 3  Vector for each document
        rain helps crops grow need and sun
  D1     1     1     1    1    0   0   0
  D2     1     0     1    0    1   1   1

Step 4  Stack the vectors into the
        document vector table above

Name every step: the schemes give half a mark for naming and half for doing. Keep the dictionary in the order words first appear, and count, do not tick, so a repeated word scores 2.

Other four-mark shapes from the recent papers

  1. Compare AI, ML and DL with one example each and draw how they nest.
    Answer AI is any technique that lets a machine act intelligently (a chess program); ML is AI that learns patterns from data (a spam filter); DL is ML using many-layered neural networks (face unlock). Draw DL inside ML inside AI. One mark per term, one for the labelled diagram.
  2. Name the AI domain for each, with a reason: (i) an app that reads a shop sign aloud in another language; (ii) a camera that counts cars at a toll gate.
    Answer (i) Natural Language Processing, because it translates text, though reading the sign first uses vision, so mention both for safety; (ii) Computer Vision, because it detects and counts objects in images. One mark for each domain, one for each reason.

Where students lost these marks in our mocks

  • No table drawn. Writing TP = 120 in a sentence instead of a labelled matrix.
  • Reality and prediction swapped. The 2025 scheme put prediction down the side; the 2025-26 sample paper put reality down the side. Label both axes and you are safe either way.
  • Steps not named in Bag of Words, so half the marks vanish.
  • Over-long answers. 50 to 80 words is the printed limit for four marks.

How each mark is split is set out on the 417 marking scheme, explained.

Q5. Are older papers still worth solving?

AI class 10 board paper 2024 solution and older papers: what still counts

Partly. Papers up to 2025 were set on the older syllabus, so some of their questions test topics that are now practical only or gone from the theory paper. Use them, but filter first.

Topic in older papersStatus for the 2027 theory paperWhat to do with it
Confusion matrix, precision, recall, F1Still theory, Unit 3Solve every one you find
Bag of Words, stop words, stemming, chatbotsStill theory, Unit 6Solve every one
Neural networks, supervised and unsupervised learningStill theory, Unit 2Solve, and add the rule-based against learning-based comparison
Problem scoping, the 4Ws and the problem statement templateStill theory, Unit 1, as part of revisiting the project cycleSolve, then add an ethics follow-up of your own
Data Science applications, SQL, CSV files, data sourcesUnit 4 Statistical Data, practical onlySkip for the written paper; know it for the viva
Python syntax and output questionsUnit 7 Advance Python, practical onlySkip for theory; practise for the lab exam
Multiple intelligences, IoT and general AI awarenessNot named in the 2026-27 Class 10 unitsSkip

A note on finding papers

CBSE publishes its sample papers and marking schemes on cbseacademic.nic.in, and has posted main-exam marking schemes on cbse.gov.in. Those are the safest sources. Copies on other sites can carry typing errors in the answers, so check any number you copy by working it yourself.

What we could not verify

We read one set of each paper. Other sets of the same year use different questions in the same blueprint, so a topic we list for 2026 may have been asked differently in your region's set. We have not checked any 2024 or earlier paper ourselves, and the second board exam of 2026 is not covered here.

Q6. How are past papers used in class?

How the live batch turns past papers into a routine

Solving an old paper once tells a student their score. Rehearsing its shapes with new numbers every week is what makes the real paper feel familiar.

One shape a Monday

Each Monday class takes one repeating shape from the table above, works a fresh version on the shared screen, then gives students a second version to solve against the clock while the teacher watches the working.

Marked like the examiner

Answers are marked with the half-mark splits the schemes use, out loud, so students hear why a missing formula line costs half a mark even when the final number is right.

Full papers in term two

From December, Wednesday classes become timed half papers, and full two-hour papers follow before the board exam, with the choice of questions planned in advance.

The mock calendar is on sample paper and mock tests, and the longer case-style questions have their own set on case based questions for Class 10 AI.

Batch or personalised: you choose

Practise past-paper shapes in the batch, or one to one

The same repeating shapes and the same marking, in two formats. The batch keeps a steady weekly rhythm with classmates sitting the same timed questions; one to one spends the hour on whichever shapes your child still drops marks on. Monthly billing, no admission fee, stop at the end of any month.

Option 1 · Fixed batch timings

Group batch, Monday and Wednesday 9 PM

₹1,499

per month, two live classes a week

  • Monday and Wednesday, 9:00 PM IST, live on video
  • Fifteen to twenty Class 10 AI students, one teacher
  • One repeating board shape worked each Monday
  • Timed half papers on Wednesdays from December
  • Full two-hour papers before the board exam
Book a free demo in this batch

Option 2 · Choose your own timing

Personalised one to one

₹4,999

per month, one private class a week, at a time set with you

  • Days and time chosen by you, in your own time zone
  • Only the shapes your child keeps losing marks on
  • Matrix and Bag of Words cases repeated with new data
  • Answers rewritten until they fit the word limits
  • Same CBSE 417 syllabus and teachers
Ask for a personalised plan

A student who already scores well but stalls on the four-mark cases usually needs only a few one to one sessions on those two shapes. A student starting the subject late gains more from the batch, where every unit is taught and then tested in the board format. Fees are shown in your own currency.

Studying outside India?

Previous year paper practice for CBSE students abroad

CBSE schools outside India sit the same 417 paper with the same frame, so a student in Sharjah or Kuwait faces exactly the shapes described above. These are the batch times in local clocks this winter.

Where you liveMon and Wed 9:00 PM ISTBest format
India (IST)
every Indian city		9:00 PMBatch fits
UAE and Oman
Dubai, Abu Dhabi, Sharjah, Muscat		7:30 PMBatch fits
Saudi Arabia, Qatar, Kuwait, Bahrain
Riyadh, Jeddah, Dammam, Doha, Kuwait City, Manama		6:30 PMBatch fits
Nepal
Kathmandu		9:15 PMBatch fits
Ethiopia and East Africa
Addis Ababa and East Africa Time		6:30 PMBatch fits
Singapore and Malaysia
Singapore, Kuala Lumpur		11:30 PMChoose 1-on-1

Where 9:00 PM IST falls too late, one to one runs at a local hour. More on studying from abroad is on CBSE Class 10 AI classes in the Gulf and abroad.

What families say

Rated 4.9 across 547 Google reviews

Real reviews from real families. We neither write nor commission them.

★★★★★

"The one step solution for my son. Modern Age Coders make learning coding so simple that kids love it. The teachers explain complex concepts clearly with practical exercises and interactive content."

Ria Mukherjee

Parent

★★★★★

"Modern Age Coders has been a game-changer for me. I struggled to grasp IT concepts and coding before joining, but their classes transformed everything. I can now confidently write complex programs with ease."

Samriddha Mondal

Student

★★★★★

"One of the most wonderful education centres out there. Education is not limited to school syllabus but focuses on skill development."

Vansh Agarwal

Student

★★★★★

"My child Dhairya is really enjoying the Modern Age Coders classes. This is his first online class and he eagerly looks forward to it. I can already see his improvement, and the teachers are very cooperative."

Sonam Oswal

Parent of Dhairya

★★★★★

"Modern Age Coders have wonderful teachers who teach in a clear, easy and practical way. The teacher boosts students' confidence and inspires them to learn without hesitation."

Sonu Goyal

Parent

★★★★★

"I highly recommend this computer coding class! The teachers are incredibly knowledgeable and passionate about coding."

Ritu Kedia

Parent

The rest of the Class 10 series

More pages for Class 10 board students

Every page in the CBSE Class 10 AI (417) series, grouped by what you need: classes, each unit, question practice, notes and revision, and the exam itself.

Classes, tuition and batches

Unit by unit

Questions and practice

Notes and revision

Exam, marks and practicals

Choosing the subject

ICSE Class 10 Computer Applications (Java): The Tue and Wed ICSE Java batch ICSE Computer Applications syllabus ICSE Java programs practice ICSE CA important questions

Questions about Class 10 AI past papers

What students ask about class 10 AI previous year question papers

Where can I get the class 10 AI previous year question paper?

CBSE posts sample papers and their marking schemes on cbseacademic.nic.in, and has published main-exam marking schemes on cbse.gov.in, which reproduce the questions. Those are the safest sources. This page describes what the 2025 and 2026 main papers asked and gives original practice in the same shapes, with worked answers.

What was asked in the CBSE class 10 AI question paper 2026?

In the set we read, Section A mixed MCQs, assertion-reason and two-statement items across Part A and every theory unit. Section B asked, among others, the seven Cs of communication, ICT uses, sustainable development, ethics frameworks, RGB storage, train-test split, stemming, a comparison of AI, ML and DL, an AI domain question, a cyber-security confusion matrix case and a Bag of Words table.

Is the 2025 AI paper still useful for 2027?

Partly. The 2025 paper followed the older syllabus, so its items on SQL, CSV files, data science applications, IoT and multiple intelligences are no longer theory topics. Its confusion matrix case, Bag of Words table, neural network question, chatbot item and ethics essay still match the 2026-27 units closely.

What are the most repeated questions in class 10 AI?

In every paper we read: a four-mark confusion matrix case, a four-mark Bag of Words table, a stemming or lemmatisation item, a chatbot type, a false positive or false negative from a situation, and an overfitting or unseen-test-data item. Neural networks and ethics appeared in every paper, in different mark bands.

Are assertion reason questions asked in the class 10 AI 417 paper?

Yes. The 2025 main paper, the 2025-26 sample paper and the 2026 main paper we read each had assertion-reason items in Section A. CBSE has not said how many a 2027 paper will include, so practise them as one Section A format among several.

How many questions are in the class 10 AI board paper?

Twenty-one, of which you answer 15. Section A has five objective questions worth 24 marks, each with six one-mark parts and a choice. Section B has sixteen written questions worth 26 marks, of which you answer ten. The paper lasts two hours.

Do previous year papers include Python questions?

The 2025 paper, set on the older syllabus, touched data handling topics. From 2025-26, Advance Python and Statistical Data are assessed only in the practical, so the theory paper does not ask you to write code. Practise programs for the practical file and lab exam instead.

How many previous year papers should I solve for class 10 AI?

Solve the 2025-26 sample paper and the 2026 main paper fully under time, then use the 2025 paper for the topics that still count. After that, fresh questions in the same shapes help more than further old papers, because the newest syllabus has only one board paper so far.

Is this page a copy of the CBSE question paper?

No. We describe each paper in our own words and write our own questions and answers. For the exact wording, read the official marking schemes. Our aim is to show the shapes that repeat and the working that earns each mark.

Are past papers part of your classes or extra?

Part of ordinary teaching. Repeating shapes, timed half papers and full mocks are all covered by the monthly fee shown on this page, in your own currency, for the batch or one to one. There is no admission fee and the first class is a free demo.

Start

Try one past-paper shape free with a teacher

In the free first class your child solves one four-mark case in a shape the board has repeated, then watches it marked half-mark by half-mark against the scheme. Afterwards the mentor tells you plainly which shapes are already safe, which need work, and whether the batch or one to one fits better.

Want the subject in one view first? Read Class 10 AI notes, unit wise.

WhatsApp us · +91 91233 66161 · connect@modernagecoders.com

Every class is live video. The form requests a callback and reserves nothing.

We call within a few hours. No card, no enrolment fee.

Want it sooner? Book a priority demo

A full live class of about 45 to 60 minutes at a slot you pick, with a written skill report and learning roadmap on WhatsApp afterwards. The fee is adjusted against your first month if you enrol.

Book a priority demo

Keep exploring Modern Age Coders

By class and age

Learn more

Free resources

From the blog

Start here