onrup

Model architecture

Dense model

What is dense model?

A dense model uses every parameter for every token. Compute and memory both scale with the full parameter count, which makes its behaviour simple to predict and its fine-tuning well understood.

Most models are dense, and most published fine-tuning practice assumes it. For a first fine-tune the predictability is worth something in itself.

The comparison against a sparse model of the same total size is not simple: the sparse one is faster per token, and the dense one usually holds more capability per gigabyte of memory.

Related terms

All terms in the glossary →

Start with the free tier

A magic link creates your account, your tenant and your first API key. No card until you ask for compute.