“In Cohere's on-premise deployment model, the enterprise customer is responsible for the inference costs of the AI models.”