STEP 2025: Grade Boundaries and How to Prepare

The 2025 STEP season kept the two-paper format in place since 2021: STEP 2 (paper code 9470) and STEP 3 (paper code 9475), STEP 1 having been discontinued from 2021. Each paper is a 3-hour written examination of 12 questions (8 Pure Mathematics, 2 Mechanics and 2 Probability and Statistics); every question is marked out of 20, the paper is marked out of 120, and only your best 6 answers count towards the total. Since 2024 the examination has been delivered by OCR, part of the same Cambridge group, and OCR also published the official results documentation for 2025.

The official grade boundaries were as follows (source: the official Explanation of Results for STEP 2025 published by OCR; each paper is out of 120):

There were no structural changes this year: the paper line-up, the marking scheme and the awarding body were all the same as in 2024.

STEP results are reported on five grades: S (Outstanding), 1 (Very Good), 2 (Good), 3 (Satisfactory), and U (Unclassified) below that. The scoring rules behind the boundaries are simple: each question carries 20 marks, only your best 6 questions count, the maximum is 120, and a fully correct answer earns full marks whatever method you use.

Against 2024, the two papers moved in different directions, and both years' figures here are official. On STEP 2 the grade 1 boundary was 69 in 2024 and 75 in 2025, grade 2 went from 51 to 61 and grade 3 from 26 to 35, while S edged from 93 to 92. On STEP 3 every boundary sat lower than in 2024: S at 86 against 93, grade 1 at 61 against 70, grade 2 at 49 against 57, and grade 3 at 27 against 34.

Boundaries are set afresh each year in the light of how demanding the questions turned out to be and how candidates actually performed, so year-to-year movement is normal and last year's boundary is not a target for next year. One further note: Cambridge offer conditions vary by college and by year, so always go by the wording of the offer itself.

The statistics below describe the STEP 2025 questions catalogued by FrontierVue and reflect the platform's own tagging, not an official analysis of the papers. In the database, STEP 2 and STEP 3 each have 12 questions, in both cases 8 Pure Mathematics, 2 Mechanics and 2 Probability and Statistics, matching the official paper structure. On the platform's difficulty labels, 11 of the 12 STEP 2 questions are tagged HARD and 1 MEDIUM; on STEP 3 the split is 10 HARD and 2 MEDIUM.

Across the two papers together, still on FrontierVue's tagging, the most frequent topic tags are Pure Mathematics (16) and Curve Sketching (5), followed by Induction and Recursion, Integration Techniques, STEP Probability, Coordinate Geometry and STEP Mechanics (4 each), then Complex Numbers and Inequalities (3 each) and Differential Equations (2). Read these as a map of where the year's questions clustered under the platform's tagging scheme rather than as a syllabus weighting.

The practical takeaway: question types that reward careful written argument, such as curve sketching and induction, featured prominently this year, so it pays to practise them to the standard of a full written solution rather than just reaching the answer.

Open the STEP past paper library and pick the 2025 sitting: both STEP 2 and STEP 3 can be attempted in the timed interface, with question-by-question analysis after submission and a full worked solution for every question. FrontierVue's STEP collection runs to 111 papers from 1987 to 2025, 1,532 questions in total; under the free-access policy the latest two past papers in each subject are free to attempt, and 2025 is the most recent year in the collection. The resource library also carries the official OCR specification and STEP Support Programme materials.

Pair the papers with the guides: the STEP 2 guide, the STEP 3 guide, writing solutions that score, choosing questions in the exam and how grades and boundaries work. After finishing a 2025 paper, place your mark against the boundary table on this page: judging yourself by grade band is closer to how the real examination is assessed than a raw accuracy rate.