From £99/mo. Always-ready inference. Zero data retention. Isle of Man data residency.
# Your existing OpenAI code works as-is
client = OpenAI(
base_url="https://iom.sovai.im/v1", # runs on IOM
api_key="your-sovai-key"
)
response = client.chat.completions.create(
model="llama-3.2-70b", # local or premium
messages=[{"role": "user", "content": "Hello!"}]
)These are the problems we built SovAI to solve
GPU rentals cost £1,000+/mo — but your AI only runs during business hours. You're literally paying to watch servers sleep. SovAI gives you always-ready inference from £99/mo.
Big cloud providers process your data in jurisdictions you can't control. When clients ask about compliance, you have to shrug. SovAI lets you choose exactly where your data lives.
Switching AI providers means rewriting your integration. When a better model comes out, you're stuck. SovAI uses the OpenAI API standard — switch models with one parameter change.
Getting started is simpler than you think
Get early access and founding member pricing. No sales calls, no enterprise contracts.
Swap your API URL — your existing OpenAI-compatible code just works. No rewrites needed.
Deploy AI features knowing exactly where your data lives. Switch models anytime with one parameter.
One line change. That's it.
# One line change. Choose your region and model.
client = OpenAI(
base_url="https://iom.sovai.im/v1", # API runs on IOM
api_key="your-sovai-key"
)
response = client.chat.completions.create(
model="llama-3.2-70b", # Model runs on IOM
messages=[{"role": "user", "content": "Hello!"}]
)
URL = where your request is processed. Use iom.sovai.im for Isle of Man, uk.sovai.im for UK, or eu.sovai.im for EU.
Model = where the AI runs. Choose local models (llama-3.2-70b) at 1K tokens = 1 credit, or premium models (gpt-4o, claude-sonnet, mistral-large) at scaled rates.
All processing happens on the Isle of Man. Your data never crosses international boundaries. GDPR compliant without the complexity.
We process your request, return the result, and throw everything away. Your data is never stored, logged, or used for training. What happens on-island stays nowhere.
Simple credit system. Local models: 1K tokens = 1 credit. Premium models scale proportionally. No surprise bills, no "we'll call you" pricing.
Get an API key. Point your code at us. Done. Your existing OpenAI-compatible code works instantly — no SDKs or gateway servers to install. Start building in minutes, not days.
No Azure subscriptions. No AWS IAM roles. No GCP project setup. No billing configurations, SDK installs, or credential juggling. Just an API endpoint and a key.
All plans include access to local and premium models. Use cost-effective on-island models (Llama 8B, 70B) or global models (GPT-4o, Claude, Mistral) — all through the same API. Credits scale by model cost.
Stop paying £1,000+/mo for GPUs that sit idle 80% of the time. SovAI is always ready — plans from £99/mo for inference when you need it, not for servers sleeping.
Use whatever dev and ops tools you want. You're not locked into vendor dashboards, SDKs, or monitoring stacks. It's your infrastructure, your way.
As new models emerge, we add them. One API, always current. No rewrites when the next big model drops — just change a parameter.
Trusted by organisations that can't compromise on data security
Protect citizen data with AI that meets public sector security requirements. Process sensitive information without overseas data transfers.
Meet FCA and regulatory requirements while leveraging AI for client services, analysis, and operations.
Maintain client confidentiality, patient privacy, and legal privilege. AI-powered assistance with complete data control.
Get more for less
| SovereignAI | The Others | |
|---|---|---|
| Starting Price | From £99/mo | £1,000+/mo minimum |
| Integration Complexity | One line code change | Platform SDK required |
| Vendor Lock-In | Zero - standard API | Gateway servers trap you |
| Getting Started | Early access waitlist | Sales cycles, demos |
| Data Location | Isle of Man only | Multiple jurisdictions |
| Data Retention | Zero — we throw it away | Stored for training/analytics |
| Infrastructure | Tier 3+ data center, NVIDIA GPUs | Unknown — could be a basement laptop |
| Software Install | None — just an API call | Their code running in your network |
| Model Choice | On-island + GPT-4o, Claude, Mistral | Only models they provide |
Every month without sovereign AI means:
on GPU rentals you're not fully using. That's £12,000+/year watching servers sleep.
when clients ask where their data goes. "Somewhere in the cloud" isn't a good answer.
to ecosystems that get harder to leave. The longer you wait, the more code you write that's tied to one provider.
The before and after of sovereign AI
No sales calls. No surprises. Join the waitlist for early access.
Perfect for developers and small teams exploring AI integration.
For growing businesses with production AI workloads.
For teams with high-volume AI workloads and compliance needs.
For large organisations with compliance and SLA requirements.
All plans include: Zero data retention • IOM data sovereignty • No vendor lock-in • Cancel anytime
Run local models at 1 credit per 1K tokens, or use premium models at these rates:
| SovAI Local | Credits |
|---|---|
| Llama 3.2 8B | 1 |
| Llama 3.2 70B | 1 |
| OpenAI GPT | Credits |
|---|---|
| GPT-5 Nano | 1 |
| GPT-5 Mini | 2 |
| GPT-5.2 Mini | 4 |
| GPT-5 | 8 |
| GPT-5.2 | 14 |
| OpenAI Reasoning | Credits |
|---|---|
| o4-mini | 10 |
| o3 | 15 |
| o1 | 95 |
| o3-pro | 145 |
| Anthropic | Credits |
|---|---|
| Claude Haiku | 27 |
| Claude Sonnet | 82 |
| Claude Opus | 136 |
| Mistral | Credits |
|---|---|
| Mistral Small | 2 |
| Mistral Medium | 11 |
| Mistral Large | 10 |
| Other | Credits |
|---|---|
| More coming soon... | — |
All processing happens on servers physically located on the Isle of Man. We run AI inference locally using open-source models — there are no external API calls to overseas providers. Your data enters our IOM-based infrastructure and the AI response is generated entirely on-island.
We deploy leading open-source models (Llama 3.x and similar) optimised for on-premises inference. These models are excellent for most business tasks—document processing, summarisation, analysis, customer service, and conversation.
We continuously evaluate new models and can deploy specific models based on enterprise requirements.
Our on-island models handle 80%+ of typical use cases — writing, summarising, document analysis, customer service — and you won't notice a difference from the big cloud models. Cost-effective, fast, and your data never leaves the Isle of Man.
Need more power? Just change the model parameter. Switch to uk-gpt4o or eu-claude-sonnet and we'll route your request to that foundation model in your chosen GDPR-compliant region. You decide how much power you need and where your data goes — per request if you want.
Most of your workload runs on affordable on-island models. The 20% that needs frontier capability routes to the big models. You're in control, not locked in.
We're an API, not a platform. No gateway servers, no proprietary middleware, no vendor lock-in. Change one line of code to use our OpenAI-compatible API. Other platforms require complex SDKs and trap you with proprietary systems. Plus, we're 10x cheaper — from £99/mo vs £1,000+/mo competitors.
Our servers run on Nvidia GPUs in enterprise-grade, professional Tier 3+ data centers with ISO 27001 (Information Security) and ISO 9001 (Quality Management) certifications - not a GPU in someone's basement. Our infrastructure includes encryption at rest and in transit, access controls, audit logging, and regular security assessments. We're also working toward SOC 2 Type II certification for additional assurance.
Yes. For organisations requiring complete isolation, we offer dedicated infrastructure deployments. This provides your own AI instance with dedicated compute resources, network isolation, and custom security configurations. Contact us to discuss your requirements.
Transparent, predictable pricing starting at £99/mo — 10x cheaper than competitors who charge £1,000+/mo minimum. Join the waitlist for early access — no enterprise sales cycles. You can also opt for dedicated capacity with SLA guarantees. Contact us for usage-based pricing or custom requirements.
Unlike platforms that lock you in with proprietary gateway servers, we provide a standard OpenAI-compatible API. You're not dependent on our infrastructure - switch providers with one line of code. The infrastructure serves you, not the other way around. True data sovereignty means you control where your data goes and who processes it.
Join the waitlist to be first in line. Starting at £99/mo — no sales calls, no complex platforms, just a simple API with Isle of Man data sovereignty.
Or email us directly at hello@sovereignai.im