UINT4 - 4-Bit Unsigned Integer | Integer Format Guide

Bit Layout

Value

V

4 bits

A 4-bit unsigned integer (also called a nibble) uses all 4 bits for magnitude, giving a range of 0 to 15 (2⁴ - 1). Two UINT4 values pack into a single byte, and each corresponds to one hexadecimal digit (0–F).

Overview

UINT4 is a sub-byte format primarily used as a packing container for quantized ML weights. While INT4 (signed) is more common in quantization schemes like GPTQ and AWQ, UINT4 appears in frameworks that use asymmetric quantization with a non-zero zero-point offset.

With only 16 possible values (0–15), UINT4 achieves the same 8× compression over FP32 as INT4. Two UINT4 values are packed into a single byte, with the first element in the least-significant nibble.

UINT4 in Quantization

In asymmetric quantization, the float-to-integer mapping includes a zero point:

quantized_value = round(float_value / scale) + zero_point

The zero point shifts the mapping so that zero maps to a non-zero integer, allowing the full 0–15 range to cover an asymmetric distribution of float values.

See also: INT4 For the signed variant using two's complement (range -8 to 7), which is more common in symmetric weight quantization schemes.

Range & Properties

Property	Value
Total Bits	4
Minimum Value	0
Maximum Value	15
Representable Values	16

All Representable Values

With only 16 possible bit patterns, here is the complete set:

Value	Bit Pattern	Signed	Unsigned
Zero	0000	0	0
One	0001	1	1
Max Unsigned	1111	-1	15
Max Signed	0111	7	7
All Ones (-1)	1111	-1	15
Min Signed (-8)	1000	-8	8

Interactive Bit Visualizer

With only 4 bits, you can click through every possible value. Each nibble corresponds to one hex digit.

Decimal:

Hex:

Format Comparison

Property	UINT4	UINT8	UINT16	UINT32
Total Bits	4	8	16	32
Signed Min	-8	-128	-32,768	-2,147,483,648
Signed Max	7	127	32,767	2,147,483,647
Unsigned Max	15	255	65,535	4,294,967,295
Step (ULP)	1	1	1	1
Values	16	256	65,536	4,294,967,296

Where UINT4 Is Used

ONNX packing: The ONNX specification defines UINT4 in TensorProto.DataType with a packing rule of two 4-bit elements per byte (first element in LSB). This is used for asymmetric quantization with unsigned zero-point offsets.
Python libraries: The ml_dtypes library provides uint4 as a narrow integer NumPy extension (4-bit, stored in a byte, with defined cast behavior).
GPU hardware: The PTX ISA documents .u4 packed types for unsigned 4-bit MMA instructions on Turing/Ampere. The MLIR AMDGPU dialect supports unsigned i4 via udot8 instructions on gfx906+.
Hexadecimal representation: Each UINT4 value corresponds to exactly one hex digit (0–F), making UINT4 the natural unit for displaying and parsing hexadecimal data.