You'd have to bench it, but the only thing I could think of that'd be faster is the standard +num coercion that's also used in the code snippet. Unfortunately,

You'd have to bench it, but the only thing I could think of that'd be faster is the standard +num coercion that's also used in the code snippet. Unfortunately, that does generate NaNs. Edit: Double checked the spec (it's quite readable) and the main difference is that + will coerce ignoring whitespace while parseFloat will ignore leading whitespace and parse leading digits. E.g. var x = '1.23foo'; parseFloat(x) is 1.23; +x is NaN. Neither looks like it should be faster than the other.