# What Happens When You Chat with an LLM?

You type a question. A model answers one token at a time. Under that: TLS, gateways, prompt assembly, tokenization, transformer layers, KV cache, GPU kernels, batching, routing, streaming, and…