jq merge values of same key into an array?

I have no idea how you'd do it with jq (I find it easier to write something in a general-purpose language like perl than to learn single-purpose tools like jq...if I used json data more often, it would probably be worth my time to learn more than its most basic uses), but it's pretty easy with perl's JSON module.

#!/usr/bin/perl
use strict;
use JSON;

# hash-of-arrays to hold an array of "level"s for each "subject"
my %merged;

# slurp the input into scalar string variable $text
my $text; { local $/; $text=<> }

my $json = JSON->new->allow_nonref;
my $data = $json->decode($text);  # parse json text into $data arrayref

# loop over each row of $data and insert into %merged
foreach my $row (@$data) {
  my $key = $row->{subject};
  my $val = $row->{level};

  push @{ $merged{$key} }, $val;
};

# pretty-print %merged as json
print $json->pretty->encode(\%merged), "\n";

sample output:

$ ./merge-json.pl input.json 
{
   "Accounting" : [
      "A Level",
      "IB",
      "University",
      "GCSE",
      "Mentoring",
      "13 Plus",
      "11 Plus"
   ],
   "Foo" : [
      "Foo Intro",
      "Foo Basics",
      "Intermediate Foo"
   ]
}

If you're running Debian or Ubuntu or similar, you can install the JSON module with sudo apt-get install libjson-perl libjson-xs-perl (this should install both the perl native version and the faster compiled C version of the module, JSON::XS). It's probably packaged for other distros too. Otherwise, install it with cpan.

A different approach from that of steeldriver's is to group the data by the subject key, then collect all the level values as arrays under the respective subject.

The following uses the same test data that steeldriver is using:

$ jq 'group_by(.subject) | map({ (first.subject):  map(.level) }) | add' file
{
  "Accounting": [
    "A Level",
    "University",
    "Mentoring",
    "13 Plus"
  ],
  "Foo": [
    "IB",
    "GCSE",
    "11 Plus"
  ]
}

The add at the end merges the elements of the array that the preceding map() creates.