Multi-Byte Integers in Hex: Endianness for Binary Formats

Single bytes in hex are easy: 4A is one value. Trouble starts at 16-bit, 32-bit, and 64-bit integers written as hex. The same digits mean different numbers depending on which byte is first—endianness.

IP addresses are one special case (IPv4/IPv6 hex). This article covers the general integer case you meet in file formats, binary protocols, and language pack/struct APIs.

Big-endian vs little-endian

Number: 0x12345678

Big-endian bytes:    12 34 56 78
Little-endian bytes: 78 56 34 12

Big-endian (network order): most significant byte first. Common on the wire.
Little-endian: least significant byte first. Common on x86/x64 hosts.

If you reverse the byte order by mistake, you still get a “valid” integer—just the wrong one.

Reading widths from a dump

Know the field width (u16 / u32 / u64).
Know the endianness from the format spec.
Slice that many bytes from the correct offset.
Assemble with shifts or a structured unpacker.

import struct

data = bytes.fromhex("12345678")
print(struct.unpack(">I", data)[0])  # 305419896 big-endian
print(struct.unpack("<I", data)[0])  # 2018915346 little-endian

<?php
$bin = hex2bin("12345678");
$be = unpack("N", $bin)[1]; // unsigned long big-endian
$le = unpack("V", $bin)[1]; // unsigned long little-endian

PHP pack/unpack letters matter—see pack() guide.

Where teams get hurt

Copying hex from a debugger that already displays host endianness, then treating it as wire order.
Documenting examples without stating endianness.
Mixing uint32_t casts with byte arrays on different CPUs.
Assuming “hex order on screen” equals “memory order.”

Practical habit

Write the endianness next to every multi-byte hex example in your docs:

length = 0x00000100 (u32 BE) → bytes 00 00 01 00
length = 0x00000100 (u32 LE) → bytes 00 01 00 00

Floating point and other layouts

IEEE-754 floats also have endian concerns when stored as raw bytes. Do not reinterpret float hex with integer rules. Use language APIs that pack floats explicitly.

Checklist before you ship a binary parser

Unit-test both endiannesses with known vectors.
Reject buffers shorter than the field width.
Log fields as value=0x... (bytes=...) during bring-up.
Freeze endianness in the spec; do not “auto-detect” without a magic discriminator.

def u32_be(b: bytes, off: int = 0) -> int:
    return int.from_bytes(b[off:off+4], "big")

def u32_le(b: bytes, off: int = 0) -> int:
    return int.from_bytes(b[off:off+4], "little")

FAQ

Is network order always big-endian?
For classic Internet protocols, yes. Always read the RFC for custom protocols.

Does hex itself have endianness?
Hex text is just digits. Endianness appears when those digits represent a multi-byte integer layout in memory or on the wire.

How do I convert quickly while debugging?
Use a structured unpack in a REPL, or break the hex into bytes and reverse groups of 2/4/8 digits as needed. Hex Calculator helps with related arithmetic.