diff --git a/compiler/builtins/docs/Num.roc b/compiler/builtins/docs/Num.roc
index 813fe34e23..419672c852 100644
--- a/compiler/builtins/docs/Num.roc
+++ b/compiler/builtins/docs/Num.roc
@@ -533,7 +533,10 @@ mul : Num range, Num range -> Num range
 
 ## Convert
 
-## Convert a number to a string, formatted as the traditional base 10 (decimal).
+## Convert a number to a [Str].
+##
+## This is the same as calling `Num.format {}` - so for more details on
+## exact formatting, see [Num.format].
 ##
 ## >>> Num.toStr 42
 ##
@@ -546,6 +549,68 @@ mul : Num range, Num range -> Num range
 ## For other bases see #toHexStr, #toOctalStr, and #toBinaryStr.
 toStr : Num * -> Str
 
+## Convert a number into a [Str], formatted with the given options.
+##
+## Default options:
+## * `base: Decimal`
+## * `notation: Standard`
+## * `decimalMark: HideForIntegers "."`
+## * `decimalDigits: { min: 0, max: All }`
+## * `minIntDigits: 1`
+## * `wholeSep: { mark: ",", places: 3 }`
+##
+## ## Options
+##
+##
+## ### decimalMark
+##
+## * `AlwaysShow` always shows the decimal mark, no matter what.
+## * `HideForIntegers` hides the decimal mark if all the numbers after the decimal mark are 0.
+##
+## The [Str] included in either of these represents the mark itself.
+##
+## ### `decimalDigits
+##
+## With 0 decimal digits, the decimal mark will still be rendered if
+## `decimalMark` is set to `AlwaysShow`.
+##
+## If `max` is less than `min`, then first the number will be truncated to `max`
+## digits, and then zeroes will be added afterwards until it reaches `min` digits.
+##
+## >>> Num.format 1.23 { decPlaces: 0, decPointVis: AlwaysShow }
+##
+## ### minIntDigits
+##
+## If the integer portion of number is fewer than this many digits, zeroes will
+## be added in front of it until there are at least `minWholeDigits` digits.
+##
+## If this is set to zero, then numbers less than 1 will begin with `"."`
+## rather than `"0."`.
+##
+## ### wholeSep
+##
+## Examples:
+##
+## In some countries (e.g. USA and UK), a comma is used to separate thousands:
+## >>> Num.format 1_000_000 { base: Decimal, wholeSep: { mark: ",", places: 3 } }
+##
+## Sometimes when rendering bits, it's nice to group them into groups of 4:
+## >>> Num.format 1_000_000 { base: Binary, wholeSep: { mark: " ", places: 4 } }
+##
+## It's also common to render hexadecimal in groups of 2:
+## >>> Num.format 1_000_000 { base: Hexadecimal, wholeSep: { mark: " ", places: 2 } }
+format :
+    Num *,
+    {
+        base ? [ Decimal, Hexadecimal, Octal, Binary ],
+        notation ? [ Standard, Scientific ],
+        decimalMark ? [ AlwaysShow Str, HideForIntegers ],
+        decimalDigits ? { min : U16, max : [ All, Trunc U16, Round U16, Floor U16, Ceil U16 ] },
+        minWholeDigits ? U16,
+        wholeSep ? { mark : Str, places : U64 }
+    }
+    -> Str
+
 ## Round off the given float to the nearest integer.
 round : Float * -> Int *
 ceil : Float * -> Int *
diff --git a/compiler/builtins/docs/Str.roc b/compiler/builtins/docs/Str.roc
index 4789db17b2..b8a216d514 100644
--- a/compiler/builtins/docs/Str.roc
+++ b/compiler/builtins/docs/Str.roc
@@ -429,16 +429,22 @@ chompCodePoint : Str, U32 -> Result Str [ Expected [ ExactCodePoint U32 ]* Str ]
 
 ## If the string begins with digits which can represent a valid #U8, return
 ## that number along with the rest of the string after the digits.
-parseU8 : Str -> Result { val : U8, rest : Str } [ Expected [ NumU8 ]* Str ]*
-parseI8 : Str -> Result { val : I8, rest : Str } [ Expected [ NumI8 ]* Str ]*
-parseU16 : Str -> Result { val : U16, rest : Str } [ Expected [ NumU16 ]* Str ]*
-parseI16 : Str -> Result { val : I16, rest : Str } [ Expected [ NumI16 ]* Str ]*
-parseU32 : Str -> Result { val : U32, rest : Str } [ Expected [ NumU32 ]* Str ]*
-parseI32 : Str -> Result { val : I32, rest : Str } [ Expected [ NumI32 ]* Str ]*
-parseU64 : Str -> Result { val : U64, rest : Str } [ Expected [ NumU64 ]* Str ]*
-parseI64 : Str -> Result { val : I64, rest : Str } [ Expected [ NumI64 ]* Str ]*
-parseU128 : Str -> Result { val : U128, rest : Str } [ Expected [ NumU128 ]* Str ]*
-parseI128 : Str -> Result { val : I128, rest : Str } [ Expected [ NumI128 ]* Str ]*
+parseU8 : Str, NumFormat -> Result { val : U8, rest : Str } [ Expected [ NumU8 ]* Str ]*
+parseI8 : Str, NumFormat -> Result { val : I8, rest : Str } [ Expected [ NumI8 ]* Str ]*
+parseU16 : Str, NumFormat -> Result { val : U16, rest : Str } [ Expected [ NumU16 ]* Str ]*
+parseI16 : Str, NumFormat -> Result { val : I16, rest : Str } [ Expected [ NumI16 ]* Str ]*
+parseU32 : Str, NumFormat -> Result { val : U32, rest : Str } [ Expected [ NumU32 ]* Str ]*
+parseI32 : Str, NumFormat -> Result { val : I32, rest : Str } [ Expected [ NumI32 ]* Str ]*
+parseU64 : Str, NumFormat -> Result { val : U64, rest : Str } [ Expected [ NumU64 ]* Str ]*
+parseI64 : Str, NumFormat -> Result { val : I64, rest : Str } [ Expected [ NumI64 ]* Str ]*
+parseU128 : Str, NumFormat -> Result { val : U128, rest : Str } [ Expected [ NumU128 ]* Str ]*
+parseI128 : Str, NumFormat -> Result { val : I128, rest : Str } [ Expected [ NumI128 ]* Str ]*
 
-parseF64 : Str -> Result { val : U128, rest : Str } [ Expected [ NumF64 ]* Str ]*
-parseF32 : Str -> Result { val : I128, rest : Str } [ Expected [ NumF32 ]* Str ]*
+parseF64 : Str, NumFormat -> Result { val : U128, rest : Str } [ Expected [ NumF64 ]* Str ]*
+parseF32 : Str, NumFormat -> Result { val : I128, rest : Str } [ Expected [ NumF32 ]* Str ]*
+
+## TODO make this similar to the Num.format argument
+## except more flexible - e.g. the policy for whole number separators
+## might be to allow them, to require them, or to allow them only every N digits
+## (e.g. 3 for thousands, 4 for bits, 2 for hex)
+NumFormat : { }