Large Language Models are the engine behind ChatGPT, Claude, Gemini, and every AI tool blowing up your feed. Here's how they actually work -- no PhD required.
Hit "Run LLM" to watch tokens flow through transformer layers and generate a response.
An LLM is a robot student that read the entire internet and learned to predict what word comes next. You give it a sentence, and it finishes the thought -- one word at a time -- based on patterns it learned from trillions of examples.
That's it. No magic, no consciousness, no secret sauce. It's the world's most sophisticated autocomplete -- and it turns out that's powerful enough to write code, pass bar exams, and explain quantum physics to a five-year-old.
Click a card to expand the full breakdown. No jargon without explanation.
The models you've heard of and who makes them.
Powers ChatGPT, the model that started the hype
Built for safety and long-form reasoning
Google's flagship, multimodal from the start
Open-weight model anyone can download and run
French lab making small models punch above weight
Elon's AI company, trained on X/Twitter data