A large language model (LLM) is a neural network trained on hundreds of billions to trillions of words of text — books, websites, code, research papers, legal documents, and more — to learn statistical patterns of language. Through this training, LLMs develop the ability to understand context, generate coherent text, reason through problems, translate languages, write code, and answer questions across a vast range of domains.
The practical consequence for job markets is that LLMs can perform a large category of cognitive tasks that previously required human expertise: drafting professional communications, summarizing documents, conducting research, writing code, generating creative content, and answering domain-specific questions. When an LLM is given specialized training data or context (via RAG systems, fine-tuning, or detailed prompting), it can match or exceed junior human professionals in many task categories.
LLMs are probabilistic systems — they generate the most statistically likely response given their training and the input prompt. This means they are excellent at tasks where the "right answer" resembles patterns in their training data, but prone to "hallucinations" (confidently generating false information) when operating at the edges of their knowledge. Understanding this limitation is critical for anyone working with LLMs professionally.
The major LLMs as of 2026 — GPT-4o, Claude 3.5/3.7, Gemini 2.0 — are multimodal: they can process and generate text, images, code, and structured data. This multimodality has extended their reach into creative and visual domains beyond the original text-centric applications.