main.go

package main

import (
	"testing"
)

func A() {
	m := map[int][]string{}

	for i, _ := range []int{1, 2, 3} {
		var x []string
		for _, c := range []string{"a", "b", "c"} {
			x = append(x, c)
		}
		m[i] = x
	}
}

func B() {
	m := map[int][]string{}

	var x []string
	for i, _ := range []int{1, 2, 3} {
		for _, c := range []string{"a", "b", "c"} {
			x = append(x, c)
		}
		m[i] = x
		x = nil
	}
}

func BenchmarkA(b *testing.B) {
	for i := 0; i < b.N; i++ {
		A()
	}
}

func BenchmarkB(b *testing.B) {
	for i := 0; i < b.N; i++ {
		B()
	}
}

For me, on 64-bit Linux, "A" tends to run slightly faster (not that there is any real-life difference here), most of the time (note: this does not mean always).

$ go test -bench=. -benchmem 
goos: linux
goarch: amd64
pkg: test
cpu: AMD Ryzen 9 6900HX with Radeon Graphics        
BenchmarkA-16    	 1637730	       763.1 ns/op	     336 B/op	       9 allocs/op
BenchmarkB-16    	 1480376	       789.4 ns/op	     336 B/op	       9 allocs/op
PASS
ok  	test	3.981s