The GIF specification prescribes little-endian byte order for the hexadecimal numbers shown in Wikipedia’s example.
In a little-endian multi-byte value, the least significant byte comes first. Thus, the example’s logical-screen width of 3 appears as hexadecimal bytes “03 00,” and its height of 5 appears as “05 00.” Reading those pairs in the prescribed order produces the intended decimal dimensions.
This rule applies to multi-byte numeric fields such as image width, image height, screen position, and animation delay. It does not mean that every part of a GIF is numerically little-endian: text signatures such as GIF89a are ASCII bytes, color-table entries are stored as red, green, and blue byte triples, and compressed LZW codes have their own bit-packing rules.
Little-endian is also easy to confuse with bit order. Byte order describes how bytes in a multi-byte number are arranged; bit packing describes how individual compressed code bits are placed into the data stream. GIF uses both conventions, which is why a decoder must follow the format specification rather than interpret the file as ordinary text.