Floating-point constants suffixed with M become type BigFloat. Compare joker.math/e, a truncated version of Go's math.E due to forcing it into a Double, to the original value with Clojure's M suffix:
user=> joker.math/e
2.718281828459045
user=> 2.71828182845904523536028747135266249775724709369995957496696763M
2.71828182845904523536028747135266249775724709369995957496696763M
user=>
Further, BigFloats created from strings (when a constant such as 1.3M is parsed by Joker) are given a minimum precision of 53 (the same precision as a Double, aka float64) and a maximum precision based on the number of digits and the number of bits each digit represents (3.33 for decimal; 1, 3, or 4 for binary, octal, and hex).
The joker.math/precision function returns the precision for a BigFloat type, though it supports a few others (e.g. (joker.math/precision 1) returns 63N on 64-bit systems, 31N on 32-bit).
This combines to produce a fairly useful set of default behaviors:
user=> (joker.math/precision 2.71828182845904523536028747135266249775724709369995957496696763M)
208N
user=> (def c1 1.3M)
#'user/c1
user=> (def c2 0000000000001.3000000000000M)
#'user/c2
user=> (joker.math/precision c1)
53
user=> (joker.math/precision c2)
86
user=> (= c1 c2)
false
user=> (+ c1 c2)
2.600000000000000044408921M
user=>
Note the inequality of the two values: when c1 is binary-extended to match the precision of c2, it represents a slightly different value than does c1, as confirmed when adding the two values. This effect is not observed when using binary-based (binary, octal, or hexadecimal) encoding, since they encode the mantissa and thus the value precisely, which base-10 encoding (shown above) does not, due to the underlying representation using binary (rather than decimal) digits. E.g.:
user=> (= 0x1.fM 0x1.fM)
true
user=> (= 0x1.fM 0x1.f00000000000000000000000000M)
true
user=>
Finally, joker.math/set-precision can be used to make a copy of a BigFloat with the specified precision, a la Go's math/big.(*Float).SetPrec() method:
user=> (def big-e 2.71828182845904523536028747135266249775724709369995957496696763M)
#'user/big-e
user=> (joker.math/precision big-e)
208N
user=> (def truncated-e (joker.math/set-precision 53 big-e))
#'user/truncated-e
user=> truncated-e
2.718281828459045M
user=> (joker.math/precision truncated-e)
53N
user=> (joker.math/precision joker.math/e)
53N
user=>