A single API
We intend to offer one integration through which developers could select a supported model, send a request and receive its response.
Home / Inference gateway
Coming soon
Inference Gateway.
We are planning a gateway that would give developers access to AI models through a single API. Our approach is to begin with model access through third-party providers, with a longer-term plan to host selected open-source models on GPU infrastructure.
What we are planningOne integration.
One integration.
More model choice.
The aim is to make it simpler to bring AI inference into your applications as supported models become available.
Access to proprietary models
We plan to route requests for proprietary models to their providers, subject to the necessary commercial agreements and permissions. Those models would run on provider infrastructure.
Open-source models
Over time, we intend to host selected open-source models on GPU infrastructure we operate, expanding the model catalogue subject to model licences and capacity.
How it would workYour application.
Your application.
Our gateway.
Your application would send a request to the gateway, specifying a supported model. The gateway would route it to the relevant provider or a model hosted by Bench Mark, then return the response through the same API.
Coming soon
A planned service
The inference gateway is not yet available. Supported models, pricing and launch timing have not been announced. The details above describe our intended direction and may evolve before launch.