---
name: Float
kind: type
since: 0.1
stability: stable
---
64-bit IEEE 754 binary floating-point number.
Carries the usual floating-point caveats: `NaN != NaN`, denormals,
non-associative summation. Comparisons against `NaN` always return
`false`, so equality checks against possibly-`NaN` values should
either route through a refinement type or use a tolerance.
Used for confidence scores (`confidence_threshold: 0.85`), ratios,
and any continuous numeric value. Prefer `Integer` when discrete
semantics are intended.