float64

Float64 is a 64-bit floating-point numeric type that follows the IEEE 754 standard for double-precision binary floating-point format. It encodes numbers using one sign bit, 11 exponent bits with a bias of 1023, and 52 fraction bits, yielding a wide dynamic range and substantial precision.

Across most programming languages, float64 corresponds to a native double-precision float: C/C++ double, Go float64, Java double, Python's float on most implementations, and JavaScript's Number type. In data science libraries such as NumPy, "float64" is used as a standardized dtype name. A typical value range is approximately ±1.7976931348623157e308, with smallest positive normal around 2.2250738585072014e-308 and subnormals down to about 5e-324. The precision is about 15 to 17 decimal digits.

An 8-byte representation enables distinct values including ±0 and special values such as NaN (not a number) and ±Infinity. Zero can be signed, and subnormal numbers exist closer to zero when the exponent is zero. Floating-point arithmetic is subject to rounding errors, cancellation, and representation limits; the machine epsilon for float64 is about 2.220446049250313e-16.

In practice, float64 is used for scientific computing, graphics, and general numerical work where a balance of range and precision is needed. It is not exact for many decimal fractions, making decimal types preferable for currency or precise fixed-point calculations. Converting to and from strings, integers, or other types may introduce rounding or overflow. Understanding precision limits and error propagation is important when performing iterative computations.