The average SAT score, and what it actually tells you

The mean total SAT score for the class of 2025 was 1029, across 2,004,965 test takers. Section means were 521 on Evidence-Based Reading and Writing and 508 on Math, with standard deviations of 235 on the total, 121 on Reading and Writing, and 126 on Math (2025 SAT Suite Annual Report).

That is the newest full-year national figure available in August 2026. A class of 2026 report is expected around September or October 2026, and the number will move by a few points when it lands.

The class of 2025 in full

Measure	Total	ERW	Math
Mean score	1029	521	508
Standard deviation	235	121	126
Scale	400-1600	200-800	200-800

Test takers: 2,004,965. Met both College and Career Readiness Benchmarks: 39 percent. Met the Reading and Writing benchmark of 480: 64 percent. Met the Math benchmark of 530: 41 percent. Met neither: 34 percent.

The benchmarks themselves are worth stating precisely, because they are often misread as pass marks. College Board says they “predict a 75% likelihood of achieving a C or higher in related first semester, credit-bearing college courses.” No college uses them as an admissions threshold.

The distribution behind the mean

A single average hides the shape of the data. The class of 2025 distribution:

Total score band	Share of test takers
1400-1600	7%
1200-1390	18%
1000-1190	28%
800-990	30%
600-790	14%
400-590	2%

The two bands either side of the mean, 800-990 and 1000-1190, hold 58 percent of the cohort between them. One standard deviation of 235 points around the mean spans roughly 794 to 1264, which covers the bulk of the distribution but tells you little about the tails, since the distribution is not perfectly symmetric.

The mean and the median sit close together. In the SAT User Group percentile table, a total of 1020 is the 50th percentile, against a class of 2025 mean of 1029 (College Board Research). These are different measurements over different reference groups, so treat the closeness as a rough consistency check rather than an exact match. The full percentile tables show every ten point step.

Why the average is a weak comparison point

The mean is calculated over everyone who took the SAT, and who takes the SAT is not a fixed population. Three structural features distort it.

Participation rules vary by state

In some states the SAT is administered to all public school juniors during the school day under a state or district contract. In those states nearly the entire cohort is tested, including students with no plan to apply to a four-year college. In states where testing is voluntary, the tested group is smaller and self-selected toward college-bound students.

The result is that a state with near-universal testing will usually show a lower mean than a state with low voluntary participation, even if the underlying student populations were identical. State means are not comparable to each other, and the national mean is a blend of both regimes. College Board publishes state-level reports alongside the national one, and any state comparison should be read next to that state’s participation rate.

The comparison group changes with policy

Test-optional and test-required policies shift who bothers to test at all. When more colleges require scores, more students in the middle of the distribution sit the exam. When fewer do, the tested pool tilts toward students who expect a strong result. Neither shift says anything about whether students are learning more or less.

Two of the three published averages describe different groups

The annual report mean describes one graduating class. The SAT User Group percentile table is built from a rolling window of the past three graduating classes. The Nationally Representative Sample table models all US 11th and 12th graders whether or not they test. A number pulled from one and compared against another will not line up. Which table applies when is set out in SAT percentiles explained.

The ERW versus Reading and Writing labelling quirk

You will see two names for the same section, and the inconsistency comes from College Board itself.

The 2025 Annual Report labels the section ERW, short for Evidence-Based Reading and Writing. That name dates from the 2016 paper SAT. The digital test, its structure page, and current score reports call the same section Reading and Writing, usually abbreviated RW.

They are the same 200-800 section. Nothing about the reporting scale changed with the name. What did change with the move to digital is what sits underneath: subscores and cross-test scores were eliminated, and score reports now show performance across four Reading and Writing content domains and four Math content domains instead. If a source still quotes an SAT subscore such as Command of Evidence on a 1-15 scale, it is describing the retired paper test.

The section itself is 54 questions in two 32-minute modules on the digital test. Only 50 of those are scored, since each module carries two unscored pretest items. The same logic applies to Math, where 44 questions administered yield 40 scored. That is one reason a practice test raw score does not map cleanly onto a real one, as covered in the raw score conversion reference.

What to use instead of the average

For placing one score in the national field, use a percentile rather than the mean, because a percentile carries the shape of the distribution with it. The SAT percentile calculator reports both published systems and states which one it is using.

For admissions, use each college’s published middle 50 percent range instead of any national figure. A national mean has no bearing on a specific applicant pool. The score needed calculator works backwards from a target total to the section combinations that reach it.

For comparing against ACT averages, be careful. The two tests report on different scales and different populations, and the only official crosswalk is the 2018 ACT/SAT concordance, published with standard errors of plus or minus 2 ACT points and plus or minus 80 SAT points. It also concords a retired four-section ACT against the pre-digital paper SAT. A concorded national average carries all of that slack, so treat it as an approximation and check the bands shown by the ACT to SAT converter.

For tracking your own progress, compare like with like: an adaptive Bluebook result against another adaptive result, a paper practice test range against another paper range. Mixing the two produces a difference that reflects the scoring method rather than any change in performance.

A running record of the national figures and their vintages sits on the average SAT score reference page, and the provenance of every number used across this site is listed under data sources.