What's the difference between an LLM and a regular chatbot?

An LLM, or large language model, is the actual brain that powers a chatbot — the chatbot itself is just the interface you talk to. Think of it like a car engine versus the whole car. The LLM is the engine: it's the massive neural network trained on billions of pages of text to predict what word comes next in a sequence. ChatGPT, Claude, and Gemini are the full cars. They wrap that engine in a user-friendly chat window, add safety filters, and give it a personality. When you type a question into ChatGPT, you're not talking directly to the LLM. You're talking to an application that formats your request, sends it to the model, gets the raw prediction back, and then polishes it up for you. This distinction matters because it explains why the same underlying model can feel completely different depending on the wrapper. A model like GPT-4 can be used inside a helpful assistant, a coding tool, or even a customer service bot — and each one will behave differently based on how the developer set it up. I've seen people get frustrated because they tried an LLM through one app and hated it, then assumed all AI was the same. That's like driving a beat-up sedan and deciding all cars are slow. The model is the raw capability. The chatbot is the experience. Knowing this also helps you understand why some tools let you adjust things like 'temperature' or 'system prompts' — those are levers that control how the underlying LLM behaves, and they're not available in every chatbot interface. **Related**: Which LLM is best for creative writing? | Can I run an LLM on my own computer instead of using a chatbot?