Exact arithmetic on numbers too large for a normal calculator
Every calculator you have ever used stores numbers as 64-bit floating-point values, and that representation quietly rounds whole numbers above 9,007,199,254,740,991. This article explains the rounding problem, walks through a calculator that avoids it entirely, and covers each operation it offers. All arithmetic runs on your own device with no data sent anywhere.
The silent rounding problem most people never notice
Open any spreadsheet or standard calculator and type 9007199254740993. Add one. The answer should be 9007199254740994, but you will almost certainly get 9007199254740994 or even 9007199254740992, depending on how the tool rounds. No error appears. No warning pops up. The answer is simply wrong.
The cause is the IEEE 754 floating-point format that nearly all software uses. It stores numbers in a fixed number of binary digits, which gives perfect accuracy up to a certain size and then starts skipping values. Past 2 to the power of 53 minus 1 the gaps between representable numbers grow larger than 1, so the result is forced to the nearest available value. Financial systems, cryptography, and any field that deals in large exact counts all hit this wall.
This calculator uses big integer arithmetic instead. There is no floating-point layer at all. The numbers it handles are stored as sequences of digits with no upper limit, so the answer is exact no matter how many digits are involved.
Entering numbers and cleaning up formatting
The field labelled First number accepts a whole number of any length. You can paste a number that contains commas, spaces, or underscores and they will be ignored, so copying a formatted figure straight from a spreadsheet or a document works without editing. The same applies to the field labelled Second number.
A negative sign at the front is allowed. Leading zeros are stripped. If you paste something that contains letters or other characters that are not part of a number, the tool will tell you rather than silently dropping them.
The checkbox labelled Group digits in threes controls whether the output is printed with separator commas. It is on by default, which makes large results easier to read. Turn it off when you plan to copy the result into code or another tool that would choke on the commas.
The ten operations and how they differ
The dropdown labelled Operation lists ten choices. Add, Subtract and Multiply work on both numbers in the obvious way. These three operations are the ones you are most likely to reach for when verifying a spreadsheet calculation that looks suspicious or when combining counts that overflow a normal number type.
Divide (whole) gives the integer quotient with no fractional part, because producing a decimal would require rounding somewhere, which defeats the purpose. Remainder gives the leftover after that division. Together, they express the exact answer to any division problem without losing a single digit.
To the power of raises the first number to the exponent given by the second. This can produce enormous results very quickly. The tool estimates the size before it starts, and if the answer would have more digits than your browser can handle, it refuses rather than freezing your tab. That safety check means you cannot accidentally ask for something like 9 to the power of 9 to the power of 9, which has more digits than there are atoms in the observable universe.
GCD finds the greatest common divisor of the two numbers, and LCM finds the least common multiple. These are useful in number theory, cryptography key generation, and simplifying fractions to their lowest terms. Factorial of the first ignores the second number entirely and computes the product of every integer from 1 up to the first number. Integer square root of the first also ignores the second and gives the largest whole number whose square does not exceed the input, along with a note about whether the input was a perfect square and what the remainder was.
A worked example with the demo button
The button labelled Show the rounding bug loads a pair of numbers that demonstrate the floating-point problem. It fills in 9007199254740993 and 1, sets the operation to Add, and shows you the correct answer of 9007199254740994. Below the result a line reports the number of digits and a note that this value is beyond what a standard number type can hold exactly.
Try the same addition on any other calculator or in a browser console by typing 9007199254740993 + 1. You will get the wrong answer. That single comparison makes the problem tangible in a way that no amount of explanation can match.
For a larger example, try computing the factorial of 100. The result has 158 digits and is far too large for any 64-bit representation. Copy it into a spreadsheet and the spreadsheet will show the first few digits followed by zeros, because it cannot store the full value. The big number calculator shows every digit.
Downloading and copying the result
The result appears in a read-only text area. The button labelled Copy puts it on your clipboard. The button labelled Download .txt saves it as a plain text file, which is useful when the result runs to thousands of digits and you need to keep it or feed it into another program. When the digit grouping checkbox is on, the copied text includes commas; turn it off before copying if the receiving program does not understand grouped numbers.
The digit count shown underneath the result area is helpful for sanity checks. If you are computing an RSA modulus and expect a 2048-bit number, you know the decimal form should have about 617 digits. A glance at the count tells you whether you entered the right inputs. A note also appears whenever the result exceeds the range that a 64-bit float could hold exactly, which is the point where every other calculator would start rounding.
Why there are no decimals and how to work around it
The calculator handles whole numbers only. That is a design choice, not a limitation of the technology. Exact decimal arithmetic is possible but introduces questions about how many decimal places to keep, and different answers suit different problems. By sticking to integers the tool stays simple and the results stay beyond dispute.
The practical workaround is to scale everything to the smallest unit. If you are working with money in dollars and cents, multiply by 100 and work in cents. If you need three decimal places of precision, multiply by 1000. Do the arithmetic on those scaled-up integers and divide at the end. This is how banking software handles money internally for the same reason: floating-point dollars lose pennies on large sums.
Division with a remainder is the companion to this approach. Dividing 1000 by 3 gives a quotient of 333 and a remainder of 1, which is the exact answer rather than the endlessly repeating 333.333 that floating-point would truncate somewhere.
Common mistakes and edge cases
The most common error is pasting a number that contains a decimal point. Because the calculator is integer-only, a decimal point makes the input invalid. Remove it or scale the number first.
Another trap is trying an exponent that is itself enormous. Raising 2 to the power of a million is fine and produces a number with about 301,030 digits. But raising 999 to the power of 999,999 takes real time and memory. The pre-check catches the extreme cases, but borderline ones can still take a minute. If the browser feels sluggish, refresh the page.
Negative numbers work with addition, subtraction and multiplication as expected. Division and remainder follow the convention that the remainder has the same sign as the dividend, which matches most programming languages. Factorial and integer square root are only available for positive numbers, and the tool will tell you if you try a negative input for either one.
Privacy and browser limits
Every operation runs in your browser using the built-in big integer support that modern browsers provide. No numbers are sent to any server, no account is needed, and nothing is stored after you leave the page.
The only real limit is the memory available to your browser tab, which in practice means numbers with a hundred thousand digits or more are comfortable. Beyond that, patience is the constraint rather than correctness. A multiplication of two ten-thousand-digit numbers completes in well under a second on a modern laptop. A factorial of ten thousand takes a bit longer but still finishes, and the result has over thirty-five thousand digits.