An AI model is the foundation from which ChatGPT, Gemini and other chatbot applications are created. Models are pattern recognition systems that are fine-tuned to answer questions and generate original content.
The AI programmer designs the model, which is a "neural network" with interconnections between each node called "parameters." These parameters start out with preset values and are modified in the training stages to generate better results. Large language models (LLMs) can have billions of parameters and small language models (SLMs) have hundreds of thousands or millions.
After the model's neural network is designed, it is trained using examples in huge datasets extracted from every source on the Internet. During training, the parameters are adjusted to improve the results. For more details, see
AI training vs. inference,
deep learning and
neural network.