The LLM Lab

FreeNo sign-up

The fastest way to understand a language model is to poke at one. These small interactive tools let you see the ideas from the lessons working.

• Tokenizer playground

  Type anything and see how a real tokenizer splits it into tokens. Find out why “strawberry” is awkward for a model.

• Temperature lab

  Slide the temperature and watch the next-word probabilities reshape, then sample from them.

• Context window lab

  Shrink the window and watch your first instruction fall out of the model’s view.

• Embeddings map

  Explore real embeddings: pick a word and see its nearest neighbours in meaning, drawn on a map.

• RAG builder

  Choose which passages to hand an assistant, and see why retrieval quality decides the answer.

• Spot the hallucination

  Four AI-style answers, each with fabricated details hiding among true ones. Can you find them?

• Prompt workbench

  Write a prompt, see what a real model does with it, and get a critique of what your prompt left out. Free to try; needs an account.