Primitive Types

Nil: 0xc0 (one byte).

Boolean: 0xc2 (false), 0xc3 (true).

Integers:

• positive fixint (0xxxxxxx): 0..127 inline (1 byte total).
• negative fixint (111xxxxx): -32..-1 inline (1 byte).
• uint 8: 0xcc, value byte.
• uint 16: 0xcd, 2 BE bytes.
• uint 32: 0xce, 4 BE bytes.
• uint 64: 0xcf, 8 BE bytes.
• int 8: 0xd0, value byte.
• int 16: 0xd1, 2 BE bytes.
• int 32: 0xd2, 4 BE bytes.
• int 64: 0xd3, 8 BE bytes.

Encoder picks smallest representation:

Float 32: 0xca, 4 BE bytes (IEEE 754). Float 64: 0xcb, 8 BE bytes (IEEE 754).

Default: encoders use float64 for Python floats (preserve precision).

Strings:

• fixstr (101xxxxx): 0-31 bytes, length in low 5 bits.
• str 8: 0xd9, 1-byte length (up to 255).
• str 16: 0xda, 2-byte length.
• str 32: 0xdb, 4-byte length.
• Then UTF-8 bytes.

Binary (raw bytes, distinct from string):

• bin 8: 0xc4, 1-byte length.
• bin 16: 0xc5, 2-byte length.
• bin 32: 0xc6, 4-byte length.

Pre-2013 spec didn't have separate bin; everything was raw. Modern spec separates.