One platform.
Your kind of control.
From your first API call to dedicated infrastructure. Find the deployment that fits your workload.
Token API Services
Access open-weight models using an OpenAI-compatible API. Pay for the tokens you use, with serving in your chosen region.
- A familiar integration for your existing tools
- Open-weight model access
- A starting point for new workloads
Dedicated Endpoints
Reserve capacity for a catalogue model or your own weights. Agree your region, deployment requirements and availability terms with the team.
- Reserved capacity for your model
- Planning around your workload and demand
- Direct support with deployment requirements
Private Nodes
Discuss dedicated nodes, private networking and a named data centre, shaped around your requirements.
- Dedicated resource options
- Private networking requirements
- Location and operational needs considered together
Where it runs matters.
Bring your UK or EU location, residency and networking requirements into the deployment conversation from day one.
Explore regional routingBefore you build.
Talk to the teamWhich deployment is right for me?
Token API Services are a starting point for applications using a familiar API. Dedicated Endpoints focus on reserved model capacity. Private Nodes suit organisations that need physically dedicated infrastructure and private networking.
Can I use my existing OpenAI SDK?
InferFoundry uses an OpenAI-compatible API. Your integration needs the base URL, API key and model identifier supplied with your deployment. The homepage includes Python, TypeScript and cURL examples.
How do I choose a region?
Discuss your UK or EU location and residency requirements with the team. Available deployments and operational details will be agreed as part of that conversation.
Which models and pricing are available?
Contact us for the current model catalogue, availability and pricing. These depend on your deployment and capacity requirements.
How do I get started?
Email hello@inferfoundry.ai with what you’re building, your preferred region and expected workload.