Transformer
Definition
The transformer is a network architecture introduced by Google in 2017 whose core mechanism, "self-attention", allows the model to capture relationships between any positions in an input sequence in parallel. It is the technical foundation of practically all modern language models.
Meaning in a Business Context
Before the transformer, recurrent networks (RNN, LSTM) dominated; they processed sequences step by step and struggled with long dependencies. Transformers work in parallel and scale extremely efficiently — which is why models such as GPT, Claude, Llama and Mistral could be trained in the first place. For companies, this architecture is mainly relevant indirectly: practically every language AI used today is based on it. The architecture has a direct impact on topics such as the context window (how much text the model can process at once), latency (response speed) and cost (token-based pricing).