bit

Definition

The smallest unit of information, a single binary digit. In current practice the word appears mostly in quantization: how many bits each weight or activation is stored in — 8-bit, 4-bit, and lower — which directly sets memory footprint and therefore what fits on a given GPU. Fewer bits is not uniformly worse; degradation depends on which tensors are quantized and on the task, so bit-width claims should be read next to the evaluation.

Explainers using this term

• Paper Explained: Why Gated DeltaNet Survives 4-Bit Quantization — NVFP4 W4A4 in a Hybrid 27BWhy Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM