Model Vault lets a business use Cohere's models on dedicated infrastructure while Cohere operates the service. The launch announcement described an isolated deployment with managed model serving and scaling.
The documentation makes the division of responsibility concrete: the customer's serving infrastructure and accelerators are dedicated to that customer, while Cohere handles maintenance, updates and scaling. Applications send requests to the vault's API endpoint. The benefit is dedicated model capacity without requiring the customer to operate the serving software themselves.
Terms in this piece · Glossary
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
