The models behind our products
Our products use leading large language models from established providers, accessed through a controlled gateway.
Our approach
Rather than tie the products to a single model, we use leading large language models from established providers and choose the model that suits each task, such as understanding a job description, giving interview feedback or answering a question from a book. As better models become available, the models used by a product may change.
A controlled gateway
Products do not call model providers directly. Requests pass through a controlled gateway that:
- sends only the information a feature needs;
- applies limits on how often and how much each feature can request;
- keeps provider credentials out of the products themselves.
Grounded answers
Where accuracy matters, products ground model output in your own content. The Library, for example, retrieves passages from the book first and shows the citations its answers are based on, so you can check them.
Limitations
AI output can be wrong or incomplete. Treat matches, feedback and answers as a starting point and review them before relying on them. See the Terms of Service.
Your data
How content is shared with model providers is described in the Privacy Policy.
Programmatic access
Enterprise customers can request access to models through the Model API.