Decoder. plain-English AI glossary

INT8

● Core

Storing numbers as 8-bit signed integers, allowing values from -128 to 127.

Think of it like

A byte — the basic unit of computer memory.

Example

Instead of storing a weight as 0.847385 (float32, 4 bytes), store it as 108 (int8, 1 byte).

How it actually works

INT8 is the easiest quantization — it's hardware-native and most frameworks support it. The tradeoff: 4x memory savings but usually 1-3% accuracy loss. Good for dense models, less good for outlier-heavy models (like transformers).

For product teams

Lowest-hanging fruit for model compression.

For engineers

Use post-training int8 quantization from frameworks. Measure accuracy before deploying.

Quantization level.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open Decoder Add to Chrome