None is totally reliable. If you are absolutely sure that there are no empty lines and no multi-line sequences you can use wc -l and divide by 4, it's good enough for a quick check if nothing else is installed on a node. -l counts the lines, while -c counts bytes or -m for characters (not useful for getting read counts).

I would rather use a sequence toolkit, e.g. seqkit, with its stats command I found it to be very useful and fast to get some basic statistics. If you do QC of sequencing data as one should, fastQC will also tell you the number of reads anyway. I haven't checked how this works out with multi-line fastq but it should be fine and they are not common at all.