UK-based AI inference
AI inference that stays in the UK.
Run open models on dedicated GPUs in London, behind a private, OpenAI-compatible endpoint. Prompts, context and answers stay on UK infrastructure, run by a UK company.
- Compute region by default
- London
- GPUs, never shared with another tenant
- Dedicated
- Encrypted KV cache, per workspace
- AES-256
- Operator registered in England and Wales
- UK Ltd
What you get
- A private endpoint for any model in the catalog, or your own weights from Hugging Face, on a GPU sized for it automatically.
- UK residency by policy, not just by default: the KV cache, offloaded context and fine-tuned weights stay in the region you set.
- An OpenAI-compatible API, so existing SDKs and tools work by changing the base URL and key.
- Governed inference: per-tenant encryption, provable erasure and an exportable audit log.
- Hourly billing with no seats and no commitment. The price you approve is the price on the invoice.
Why UK teams choose UK-based inference
Sending personal data outside the UK is a restricted transfer under UK GDPR, which needs adequacy or safeguards and a risk assessment. Many public sector, legal, healthcare and financial buyers also require UK hosting outright. Running inference in London removes the question for the whole inference path, and keeps latency low for UK users.
Built for regulated workloads
| Requirement | How kvrun meets it |
|---|---|
| Data residency | London compute region by default, enforced by policy |
| Isolation | Dedicated GPU per deployment; cached context never shared across tenants |
| Security of processing | KV blocks sealed with AES-256-GCM under per-session keys |
| Right to erasure | Session keys destroyed on request, with an erasure certificate |
| Accountability | Append-only audit log mapped to GDPR, SOC 2 and ISO 27001 controls |
| Model training | Your data trains your model only, never anyone else's |
Open models, ready to deploy
Qwen, Gemma, Mistral and DeepSeek families for text, code and vision, each sized against the GPUs we run. See the model catalog or bring your own weights.
Questions
- Where are kvrun's GPUs?
- Deployments run in a London compute region by default. Ask us for the region of every component that touches your data before you sign.
- Is kvrun a UK company?
- Yes. kvrun is operated by KVRUN LTD, registered in England and Wales under company number 17504024.
- Can I fine-tune in the UK too?
- Yes. LoRA and QLoRA runs use the same UK infrastructure, and the tuned model serves from the same region. See fine-tuning in the UK.
- Does UK hosting make me UK GDPR compliant?
- It removes transfer issues for inference, but you still need a lawful basis, processor terms and retention rules. Our UK GDPR checklist covers the rest.