Models
Cortex turns the model into a setting and preserves the context with every change of provider.
Choose your model
Use your key
Determine your deployment
Switch without loss
A platform that is bound to one provider’s model binds the company’s knowledge to it. Cortex separates the model from the understanding a company has built with its system. The company chooses the model, Azure, OpenAI or Anthropic, through the Cortex contract or with its own key, the context remains.
The model as a setting
No provider is the best forever. Cortex speaks to Azure, OpenAI and Anthropic through a layer that turns switching into configuration. A new model is a decision, not a project.
Your own contract
Anyone who accesses a model through an intermediary relinquishes control over contract, terms and data path. Cortex uses the model through the Cortex contract or with the company’s own key, the contract with the provider stays where it is. Terms and data path remain the company’s concern.
On-premises or in the cloud
For some data the cloud is not an option, for other data it is the only sensible one. Cortex runs on-premises, in the private cloud or with the provider, and the choice can differ per data class. Deployment follows the data, not the other way around.
The context remains
In most systems, a model switch costs the memory. Cortex does not keep the context in the model but preserves it as conditions, restores it on the new model and checks whether the restoration is correct before it applies. The understanding a company has built switches along with it.
Data follows the policies
The choice of provider says nothing about which data it should see. Cortex decides per data class which data a model sees, at the system boundary, before the call takes place. What must stay in the company stays in the company, regardless of where the model runs.