G-Research Online Assessment: The Quant Quiz, Explained

The G-Research online assessment is a timed quant quiz sent after CV screening — most commonly reported as 10 multiple-choice questions covering probability, statistics, linear algebra, and calculus — with an ML-specific variant for machine-learning candidates and an automated coding test for engineers.

Updated August 16, 2026 · QuantVault

G-Research is a London-based quantitative research and technology firm that applies machine learning and statistical methods to financial markets, and it is consistently described as one of the highest-paying quant employers in the UK. Its screening quiz has enough of a reputation that people search for it by name — “the G-Research quiz” — but that name actually covers two different things: a private hiring assessment and a public puzzle competition. This page covers the hiring assessment: what each track sits, what the quiz tests, and how to prepare. For the interview rounds that follow, see our full G-Research interview questions guide.

Where the OA sits in the process

G-Research's own career guidance describes a four-phase process for quantitative researcher roles, and candidate reports on Glassdoor and QuantNet are broadly consistent with it:

Stage	Format	What it tests
1. Application	CV review, typically ~1 week	Research background, maths/CS pedigree
2. Online assessment	Quant aptitude quiz, ML quiz, or coding test by track	See the per-track breakdown below
3. Technical interviews	General track: four one-hour interviews; ML track: two	One deep mathematics round; stats, ML, programming, finance
4. Leadership interviews	Conversations with senior staff	Fit, motivation, research judgment

Glassdoor aggregates put the overall interview difficulty around 3.5 out of 5 with an average process of roughly three weeks — demanding, but rated below the very hardest names on our quant firms ranked by interview difficulty list. The OA is the first real filter: it comes before any interviewer time, and candidates report the pass bar is high.

What's in the OA, by track

G-Research states that quant applicants sit one of two quiz versions depending on their background — a general quantitative aptitude assessment or an ML-specific quiz — while engineering candidates get a coding screen instead. Here is what each looks like, separating what the firm says from what candidates report.

Quantitative Research — the general quant quiz

This is the assessment most people mean by the G-Research online assessment. Candidates most consistently report 10 multiple-choice questions with 5 options each and roughly a 90-minute limit — about nine minutes per question, which tells you these are derivation problems, not mental-math sprints. Reported topics include probability and stochastic processes (variants of classic setups like gambler's ruin come up), statistics with OLS regression named repeatedly, linear algebra, calculus and differential equations — unusual among quant OAs — plus light programming and finance. Some accounts describe shorter, faster quiz formats closer to 10–15 questions in 20–30 minutes, so treat any specific count and timing as indicative: the firm changes details between cycles, and post-2024 reports are thinner because G-Research holds candidates to fairly strict confidentiality.

One useful calibration point is official: G-Research has published a sample quant exam, structured as a Part A that everyone attempts in full (eight questions) plus a choice of one longer Part B question. The sample skews toward exactly the probability-statistics-calculus mix candidates describe, so if you work through it and feel comfortable, you are in the right register.

Machine Learning — the ML-specific quiz

Candidates from ML backgrounds sit a variant the firm describes as ML-specific. Expect the same mathematical core with a heavier tilt toward statistical learning: estimator properties, bias–variance, regularisation, and model-validation reasoning of the kind covered in our machine learning question bank. G-Research's own prep guidance recommends The Elements of Statistical Learning for this track, which is a good signal of the theoretical level intended — this is not an applied Kaggle screen.

Engineering — the coding test

Software and infrastructure candidates report an automated online coding test rather than the maths quiz: a small number of data-structures-and-algorithms problems, with SQL questions appearing in some accounts. The firm's recommended practice sites for this track — Codility, Project Euler, TopCoder — suggest standard DSA fluency under a clock rather than competitive-programming extremes. Our coding question bank covers the reported level. If you are new to how these screens work in general, our primer on what a quant online assessment is covers the common formats and proctoring norms.

What the quiz is actually testing

With nine minutes per question, the general quiz is not testing speed arithmetic — it is testing whether you can carry a single clean derivation through accurately, under mild pressure, and pick the right option among five plausible ones. A representative practice problem in that register (ours, not a leaked G-Research question): flip a fair coin until two consecutive heads appear; find the expected number of flips. Conditioning on the first flips gives the linear system $E = \frac{1}{2}(E+1) + \frac{1}{2}\left(\frac{1}{2}(E+2) + \frac{1}{2}\cdot 2\right)$, which solves to $E = 6$. Nothing exotic — but under a clock, with distractor options like 4 and 8 waiting, the filter is whether your setup and algebra are exactly right the first time. Multiple-choice with five options also makes triage a scored skill: banking the questions you can finish beats half-solving everything.

The public weekly quiz is not the OA

G-Research also runs public weekly and monthly puzzle competitions, and their fame is why “g research quiz” searches spike around each edition. To be clear: the public quiz is a marketing and community competition that anyone can enter; the online assessment is a private timed test sent to applicants after CV screening. Doing well on the public quiz is not part of the formal process, though it is genuinely good practice for the style of mathematical thinking the firm likes. We cover the public competition — its format, difficulty curve, and a practice plan — separately in our G-Research quiz guide.

How to prepare

G-Research's own guidance recommends 150 Most Frequently Asked Questions on Quant Interviews for the general track (we have a full guide to the Stefanica 150 questions) and Cracking the Coding Interview plus The Elements of Statistical Learning for ML candidates. A plan mapped to the reported quiz content:

Drill probability and expectation first. They are the quiz's centre of mass — work our probability question bank and expected value problems in timed sets at nine minutes per question, then tighten.
Go deep on regression and inference. OLS comes up by name in multiple reports — assumptions, the normal equations, how estimates behave as the data changes. Our statistics bank maps directly onto this.
Refresh linear algebra and ODE mechanics. Eigenvalues, quadratic forms, separable and linear differential equations — the linear algebra bank covers the reported territory.
Engineers: rehearse DSA plus SQL under a clock. Codility-style constraints, clean first-pass implementations.
Work the official sample exam last. It is the only fully reliable source on the current format; use it as your dress rehearsal, not your first exposure.

Practice for the G-Research OA

The fastest way to get quiz-ready is repetition under a clock: run timed sets from our probability bank and statistics bank, take a full OA simulation to rehearse pacing and triage end-to-end, and mix in brain teasers for the lateral-thinking questions that show up late in the difficulty ramp. Then browse the full problem library — 2,800+ worked problems, around 400 of them free.

More firm guides

Frequently asked questions

What is the G-Research online assessment?

It is a timed online quant quiz sent after CV screening. G-Research says candidates sit one of two versions depending on background: a general quantitative aptitude assessment or an ML-specific quiz. Candidates most commonly report the general version as 10 multiple-choice questions with 5 options each, covering probability, statistics (especially OLS regression), linear algebra, calculus and differential equations, plus light programming and finance.

How long is the G-Research online assessment?

The most consistent candidate reports describe roughly 90 minutes for 10 multiple-choice questions — about nine minutes per question, so it rewards careful derivation rather than mental-math sprints. Some accounts describe shorter, faster quiz formats, and G-Research changes details between cycles, so treat any specific timing as indicative rather than fixed.

Is the G-Research quiz the same as the online assessment?

No. The public G-Research weekly and monthly puzzle quizzes are marketing and community competitions that anyone can enter, while the hiring assessment is a private timed quiz sent to applicants after CV screening. The public quiz is useful practice for the style of mathematical thinking the firm likes, but it is not part of the formal hiring process.

Do software engineers at G-Research get an online assessment?

Yes. Engineering candidates report an automated online coding test rather than the maths quiz, typically a small number of data-structures-and-algorithms problems, with SQL questions appearing in some reports. G-Research's own prep guidance points engineering and ML candidates at Codility, Project Euler, and TopCoder-style practice, which signals the level expected.

Practice the real thing

QuantVault has 2,800+ quant interview problems with full solutions, intuition, and hints, firm-by-firm interview funnels, and an auto-graded coding judge. Start free.

Browse problems Firm interview guides