What are Gemini models?
The Harvard Chan AI Glossary
In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.
Gemini is Google’s family of AI models. A model is the trained component that processes information and produces a response. Gemini also names an assistant application, so a person saying “I used Gemini” may be referring either to the model or to the complete product used to access it.
Model variants can support different combinations of language, images, audio, code and other tasks. Multimodal is the term for working with more than one kind of information. It does not mean that every variant accepts every file type or performs equally well across all those tasks.
Imagine an application that explains a photograph of a diagram. A suitable Gemini model may interpret the image and generate an explanation. The application still has to supply the image in an accepted form, tell the model what question to answer and present the result. If the diagram contains an important ambiguity, the explanation needs to acknowledge it rather than invent missing detail.
Developers can call Gemini models through an API, a defined method for software to send requests and receive responses. The Gemini application provides a user-facing environment around model capabilities. Google AI Studio is another environment, aimed at experimenting with models and building applications. These are access and working arrangements, not additional synonyms for the model itself.
Gemini should also be distinguished from Gemma, Google’s separate family with downloadable model weights. Weights are the numerical settings learned during training. Similar names and related research do not mean that a Gemini account and a local Gemma installation provide the same system.
Version names and supported features change over time. A glossary should therefore explain how to read the names rather than freeze a particular ranking or price. For a real comparison, identify the version, the supplied information and the surrounding tools.
Model choice and workspace choice are separate decisions. A familiar interface may hide differences in the underlying configuration, while the same model family may be available through several products. Knowing which layer is being discussed makes claims about capability easier to interpret.
Source: Google — Gemini API Models.
If you would like to find out more about AI and improve your leadership skills, consider the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.
AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.
For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.
Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy