Token API Services
An OpenAI-compatible API serving open-weight models, hosted in the UK and EU.
A shared fleet. Each request uses the model capacity it needs.
Metered access
Access open-weight models through a familiar API, served from the region you choose and billed per token.
Budgets, attribution, alerts and hard caps apply per team and per key, so your finance team can trace usage to the workload behind it.
Your data stays yours
We never train on your prompts or outputs. We don’t keep them unless you choose storage or switch on a cache.
Prefix caching runs in GPU memory, with no cache shared between customers. Ask us which optional optimisation features are available for your deployment.
Who it suits
Developers, prototyping teams and production workloads that don’t yet need dedicated capacity.
- Start without a minimum commitment
- Integrate with your existing OpenAI-compatible tools
- Move to dedicated capacity as your workload grows
What comes next
Per-request savings receipts and a batch tier for asynchronous work are on the roadmap.
Talk to us about current availability and target dates.