Java With Us

The <code>float</code> Primitive in Java — 32-bit Floating Point

Updated on 2026-04-21 Reviewed by the Java With Us editorial team

float is a 32-bit IEEE 754 floating-point value — about 7 significant decimal digits, range roughly ±3.4×10³⁸. It's half the memory of a double but also half the precision. Use float only when memory is scarce (large arrays in graphics or audio code). Otherwise stick with double.

The f suffix is mandatory

float pi = 3.14;           // ❌ compile error — 3.14 is a double
float pi = 3.14f;          // ✅ f suffix
float pi = (float) 3.14;   // ✅ explicit cast

Precision

float  f = 0.1f + 0.2f;    // 0.3f — but noisy
double d = 0.1  + 0.2;     // 0.30000000000000004

System.out.println(1234567890f);  // 1.23456794E9 — precision lost past 7 digits

When to actually pick float

  • Very large numeric arrays where 4 bytes/element × millions matters (graphics, ML feature tensors).
  • APIs that require it (OpenGL, LibGDX, some java.awt signatures).
  • Network or file formats that specify 32-bit floats.

For everything else, double is faster on modern CPUs (no conversion overhead) and avoids a whole class of precision bugs.

NaN and Infinity

Same special values as double: Float.NaN, Float.POSITIVE_INFINITY, Float.isNaN(), Float.isFinite().

Common mistakes

  • Forgetting the f suffix — "incompatible types: possible lossy conversion from double to float".
  • Using float for money — worse than double. Use BigDecimal.
  • Mixing float and double — the computation happens in double and narrows back. Pick one and stick with it.

Related

Pillar: Java primitives. Sibling: double (almost always the better choice).

Related articles