Type any number, choose how precise it should be, and pick a rounding rule. The number is handled as exact decimal digits, so 2.675 really does round to 2.68, not the 2.67 many apps show.
| Rule | Result |
|---|
Rounding keeps the digits you need and adjusts the last one kept. To round 3.14159 to 2 decimal places, look at the first digit you’re dropping (1): it’s under 5, so the result is 3.14. For 2.675 the dropped part is exactly half (…5), so the rule decides: half up gives 2.68, while half to even also gives 2.68 because 8 is even.
Significant figures count from the first non-zero digit: 0.004567 to 2 significant figures is 0.0046, and 45,678 to 2 significant figures is 46,000 (4.6 × 10⁴).
Nearest multiple rounds to any step size: nearest 5 turns 17 into 15 and 18 into 20; nearest 0.05 is used for cash rounding, and nearest 0.25 for quarter-hour time billing.
Why do other calculators say 2.675 → 2.67? Computers store 2.675 as a binary fraction slightly below it (2.67499999…). This calculator works on the digits you typed, so it gives the mathematically correct answer.
Both round to the nearest value; they only differ on an exact tie (…5). Half up always moves away from zero (2.5 → 3, 3.5 → 4). Half to even, or banker’s rounding, goes to the even neighbour (2.5 → 2, 3.5 → 4), which avoids an upward bias when many values are added up.
Start counting at the first non-zero digit, keep that many digits, and round the rest. 0.07853 to 2 significant figures is 0.079; 123,456 to 3 significant figures is 123,000.
Look at the digit to the right of the place you’re rounding to. 347 to the nearest 10 is 350 (the 7 rounds up); to the nearest 100 it is 300 (the 4 rounds down).
Under half up (away from zero), −2.5 becomes −3. Ceiling always moves towards positive infinity, so −2.1 becomes −2; floor moves towards negative infinity, so −2.1 becomes −3.
The decimal 1.005 can’t be stored exactly in binary floating point; it is held as 1.00499999…, which rounds down. This calculator rounds the decimal digits exactly, giving 1.01.
