# What is an AI model?

A model is a set of learned numbers saved in files; an app loads those files to turn your prompt into an answer.

Sources checked: 2026-10-17

## You need a model and an app

Think of the model as the part that learned patterns, and the app as the part you use. Downloading a model does not install a chat app. Ollama, LM Studio and the Hugging Bay app can load supported model files.

## Five words you will see

- Weights: the learned numbers stored in the model files.
- Parameters: those numbers, counted. 1B means about one billion; it does not mean a 1 GB download.
- Quantization: storing numbers with less precision to reduce size. Q4 is a common four-bit family; the exact file size still matters.
- Context: the text a model can work with at once, including the conversation and its reply. Longer context needs more memory.
- Instruct or Chat: a model adapted to follow instructions or hold a conversation.

## The file type must match the app

GGUF is a model file format used by llama.cpp and supported by runners such as Ollama and LM Studio. Safetensors and MLX model folders need a compatible loader and may need several files. Renaming a file does not convert it.

For a first text chat, choose a complete chat model in a format your app supports. A tokenizer, image adapter or one part of a split download is not a complete model.

## Try a small, ordinary question

Ask: “Explain what a rainbow is in three short sentences.” A reply shows that the model loaded; it does not prove that every answer is correct. Check facts before relying on them.

## Sources

- [LM Studio first chat](https://lmstudio.ai/docs/app/basics)
- [Import a model into Ollama](https://docs.ollama.com/import)
- [Ollama context and memory](https://docs.ollama.com/context-length)

[Next: pick a model that fits](/learn/choose-a-model)