LLM (Large Language Model) is a type of AI trained on enormous amounts of text – books, articles, websites, code. The result is a model that understands human language, can respond to it, translate, summarise, write code, or generate content.
ChatGPT by OpenAI, Claude by Anthropic, Gemini by Google – these are all LLMs. Each has different strengths: Claude excels at long texts and safe behaviour, GPT-4 is versatile, Gemini integrates with Google.
For businesses, the key insight is that an LLM on its own is just the "engine" – the real value comes when you connect it to your own data (see RAG) or build an AI agent on top of it that performs specific tasks.
LLMs don't make decisions independently – they work based on instructions you give them (the prompt). The better the instructions, the better the results.