Goldman Sachs OA Questions 2026: Sections, Patterns, and How to Prepare

Goldman Sachs OA questions in 2026 are a HackerRank filter, not the Superday and not the CoderPad round people write about later. Engineering candidates usually see a short coding assessment. Some roles add multiple-choice questions on fundamentals or math. The mix depends on the team — engineering, strats, and campus drives are not the same test.

InterviewTruth does not have a dedicated Goldman Sachs question page yet. Use this guide for the structure, and use recent questions across companies when you want reports candidates are posting right now. Nothing below is "the paper from this month." Exact prompts rotate. The sections do not.

Goldman Sachs OA questions in 2026

Think of the OA as one of these shapes. Your invite email usually says which one you have.

Shape	What candidates describe	Who tends to get it
Two coding problems	Timed HackerRank, samples plus hidden tests	Software engineering, new grad and experienced
Coding plus MCQ	One or two coding problems, then short questions on complexity, CS basics, or probability	Some campus drives and hybrid roles
Math-heavier MCQ	Fewer coding problems, more quantitative items	Roles closer to strats than to product engineering

If you are interviewing for a software role, prepare for two coding problems and treat extra MCQs as a possible second section, not as the main event. Later rounds often move to a live CoderPad. Passing the OA means a person will watch you code. It does not mean the OA questions will be repeated out loud.

For a comparison with another timed screen, see Amazon SDE 1 OA questions 2026. Amazon adds a work-style survey. Goldman Sachs does not grade Leadership Principles in the OA.

How to spend the timer

HackerRank will not tell you that you missed a hidden test until it is too late. A practical order:

Open both problems. Pick the one whose pattern you recognize.
Restate the input and output in one sentence before you write code. Goldman prompts are sometimes worded as a business scene (trades, schedules, logs) wrapped around an ordinary algorithm.
Get a correct solution that fits the constraints. An elegant O(n log n) you cannot finish loses to a simple O(n log n) you submit.
Run the sample, then one case you invented: empty, duplicates, and the maximum value the statement allows.
Only then start the second problem. A finished first problem beats two half-written ones.

Use the language you debug fastest. Python is popular because the problems are about algorithms, not about manual memory. If your Python is rusty and your Java is not, use Java.

Coding patterns that show up

These are practice targets. They are the patterns engineering OAs at this level actually resolve to, once you translate the story.

Hash maps and arrays

Counting, pairing, and "have I seen this prefix" cover a lot of prompts.

Practice shape: you are given timestamps of trades and a window length. Return how many trades fall in every window of that length, or whether any two prices within k minutes differ by at least d.

Underneath, that is a sliding window or a sorted map — not a finance problem. Strip the nouns, name the data structure, then code.

Also drill: two-sum variants, frequency maps, subarray sum equals k, and merging intervals.

Graphs and grids

BFS and DFS show up as "rooms in a building," "servers that can reach each other," or "minimum steps to a cell."

Practice until these are automatic:

Connected components
Shortest path in an unweighted graph
Cycle in a directed graph
Multi-source BFS (several starts at once)

You rarely need a weighted shortest-path algorithm on this OA. If every edge has a cost, say so explicitly and pick Dijkstra. If not, BFS.

Strings

Parsing and windows, not suffix arrays. Group anagrams, longest substring without repeating characters, and validate a pattern against a string. Read the alphabet constraint. A 26-slot array is faster to write than a hash map when the statement promises lowercase letters.

Light dynamic programming and math

Some papers include one problem that is DP or combinatorics: number of ways to climb, a knapsack-style budget, gcd, or modular arithmetic. This is the problem to attempt second if the first was a gift.

If the role is quantitative, expect more of this and less "implement BFS." Still write brute force with a clear recurrence before you optimize. A correct exponential solution with small constraints can pass. Guessing the recurrence and indexing off by one will not.

Complexity and fundamentals, when the MCQ section exists

If you get multiple choice, it is usually short. The themes:

Big-O of a snippet they show you
Stack versus queue versus heap
What a database index is for, in one sentence
Simple probability: conditional probability, expected value of a die, "you draw two cards"
Occasional bit tricks or binary representation

You cannot out-study a surprise MCQ section in a weekend. You can flip through a short complexity cheat sheet and redo ten probability questions so the notation feels familiar.

A ten-day prep plan

Days 1–4. Two timed problems a day, 70–90 minutes total, HackerRank or any plain editor. Mix: one hash map or array, one graph or string. No editorial until you have submitted or the timer ends.

Days 5–6. Five problems where the story is domain-flavored and the algorithm is not. Rewrite the prompt in your own words as "given an array, return…". That translation is the whole skill.

Day 7. One DP problem and ten MCQs on Big-O and probability, only so the second section cannot surprise you into rushing.

Days 8–9. Two full mocks at the length your invite states. If the invite is not here yet, use 90 minutes and two problems.

Day 10. Reread your bugs, not new problems. The misses that repeat — off-by-one on windows, forgetting visited sets, integer division — are the ones the hidden tests catch.

When you want a different company's recent coding reports as extra drills, Amazon and Google pages are updated from candidate posts. They are not Goldman Sachs papers. Use them as pattern practice, and do not convince yourself a Google hard problem is what this OA will ask. It usually will not.

What happens after the OA

A pass moves you to live rounds. Engineering loops often include:

A CoderPad coding round, same patterns, but you must narrate
A design or object-modeling conversation for experienced candidates
A fit / behavioral conversation with the team

The CoderPad is where silence hurts you. The OA let you think quietly. Practice saying the pattern out loud on your last two mocks. Atlassian's Karat screen is a different company and a human interviewer, but the habit is the same: talk while you code, test before you claim you are done.

For how a hiring manager hears a story, effective storytelling in hiring manager interviews transfers even though the principles are not Amazon's.

Frequently asked questions

What platform is the Goldman Sachs OA on?

Engineering candidates almost always report HackerRank. Take the test in a supported browser, disable extra extensions, and use the language runtime the platform actually has. A library that exists on your laptop and not in their Python image is a bad surprise at minute 40.

How hard are Goldman Sachs OA questions?

For software roles, plan on one easier problem and one medium. The hard part is the wording and the timer, not a research-level algorithm. Quantitative roles can lean harder on math.

Do I need to study finance?

No, not for the engineering coding section. You need to notice that "a book of trades" is an array. You do not need options pricing to pass the OA.

Are the questions the same as the on-campus drive?

Campus drives and off-campus links can use different forms, including an MCQ block. The coding patterns overlap. Prepare the patterns, then follow the section list in your own link.

Where can I see questions people posted this week?

There is no Goldman Sachs company page on InterviewTruth yet. Start at recent questions and ignore anything that is clearly a different company. Recheck the week of your test. A static list of "this month's Goldman OA" goes stale faster than the patterns do.

Final thoughts

Goldman Sachs OA questions in 2026 come down to a timed coding test you can finish, plus an optional fundamentals section you should not be shocked by. Translate the story into a pattern, submit a correct first problem, and test the edges.

Practice that way for ten days. Then read whatever candidates have posted lately on recent questions before you click the HackerRank link. Good luck.