What is Kimi?
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.
Kimi is Moonshot AI’s model-family and assistant brand. A model is the trained component that interprets information and produces responses. An assistant is the application around that capability. The same name can therefore refer to the model or the assistant used to access it.
The Kimi developer platform provides access through an API: a defined way for software to send requests and receive results. Its models can support tasks such as conversation, code generation and interpretation of visual information, depending on the version. Visual understanding means processing images to answer questions or identify relevant content; it is not a guarantee of correct interpretation.
A model does not automatically have access to every outside resource. Its application may need to supply documents or connect tools, meaning software functions for tasks such as search or file access. The official platform documentation makes this distinction between model responses and access to external information.
Imagine asking an assistant to compare two manuals. It needs the actual manuals, enough relevant material to address the question and a way to identify the sources behind its answer. A model that handles long documents still needs the right documents. Capacity and relevance are different requirements.
The context window is the amount of information a model can work with in a particular interaction. Large context capacity can be useful, but it does not demonstrate that all details are used equally well. Long-document tasks need checks for omitted qualifications and confusion between sources.
Long-document capability is worth testing directly. Ask a question whose answer depends on a small qualification deep in the document, and check whether the response preserves it.
For practical use, identify the selected model, the application and the permitted connections. Then compare performance on a task with a known standard. A brand name can help locate the documentation, but it is the observed behavior of that complete arrangement that determines whether it is useful for the work at hand.
Source: Moonshot AI — Kimi API Main Concepts.
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