Sections

Language Models

A language model continues text. Everything else, from answering a question to a working database query, comes out of that one skill, and the model is wrong for the same reason it is right.

Articles in this section cover how models work and where they can be trusted: why a confident answer can be wrong, what benchmarks really measure, how models of the same size differ from each other, and what changes when a model runs on your own hardware.

We rely on measurements, not impressions. Where there is no number, we say so.