Open-source Model
An AI model whose inner workings are shared openly, so anyone can download it, study how it works, run it themselves, and often adapt it. This is the opposite of a closed model locked inside one company's service. An AI shared this openly is called an open-source model.
Some AI models are kept private. You can use them through a company’s app or service, but you cannot look inside or run them on your own equipment. An open-source model is the opposite. Its makers release it publicly, so anyone can download it, see how it is built, and run it themselves, usually free of charge.
A rough comparison is the difference between a restaurant meal and a published recipe. The restaurant serves you a lovely dish, but the method stays in the kitchen. A published recipe hands you the ingredients and steps so you can cook it at home, change it to taste, and share your version. Open-source models are the recipe: out in the open for people to use and adapt.
This matters for a few practical reasons. Open-source models can be run privately, which appeals to people handling sensitive information, and they can be freely customised for a particular purpose through fine-tuning. They also give the wider community a way to inspect and improve how AI works.
You will often hear open-source models mentioned alongside local AI, because being able to download a large language model is what makes running one on your own device possible. If you would like to explore this world further, our sister site uk-ai.news follows developments in the open AI space.