# What is ExLlamaV2? Definition and meaning

An inference library with a flexible mixed-precision quantization format and optimized kernels.

## Definition

**ExLlamaV2**: An inference library with a flexible mixed-precision quantization format and optimized kernels.

In short: Fast inference for quantized models on consumer GPUs.

## When it is used

Running quantized models quickly on a single GPU.

## Learn more

The full guide to **[ExLlamaV2](/tools/exllamav2)** explains how it works, when to use it and how it compares with the alternatives, with examples. The [glossary](/glossary) lists every term on the site from A to Z.

Previous term: [EWC](/glossary/ewc) · Next term: [Experience replay](/glossary/experience-replay)
