Run AI on your terms.

You want to put AI into production without losing control of your data, jurisdiction or infrastructure choices.

Eldrim gives you one API for running open models across approved European infrastructure. You define the rules. We handle the complexity behind them.

Tell us what you need Start building

Sweden & EU residency
OpenAI-compatible
Multi-provider
Recorded execution routes

Early access. We confirm models, eligible capacity and service terms with you before onboarding.

AI infrastructure should follow your rules

Moving from an AI prototype to production creates questions that have little to do with the model.

Where is our data processed?
Which company operates the infrastructure?
What happens when a provider runs out of capacity?
Can we change models or providers without rebuilding our application?
What can we show security and compliance about where a request was executed?

You should not have to build an infrastructure team just to answer them.

Tell us your constraints. We help you run it.

Start with what your organisation actually needs: models, latency, volume, data residency, security and compliance requirements.

We listen, agree which requirements the proposed setup can meet, and turn the agreed execution rules into routing configuration. Your application talks to one API. Eldrim handles the infrastructure behind it.

Start with an AI Infrastructure Review

A 30–45 minute conversation about your workload and its constraints. Bring what you know; we help identify the questions that still need answers.

We follow up with a recommended model or shortlist, a residency policy, a capacity arrangement and a cost estimate. The recommendation identifies assumptions and anything that needs testing or agreement before production.

Request an AI Infrastructure Review

A simple path to production

Tell us what you need.

We listen to the technical, commercial and regulatory requirements of your workload.
Define the rules.

Together we decide where workloads may run, which providers and models are acceptable, and what needs to be recorded.
Run through Eldrim.

Your application uses one OpenAI-compatible API while Eldrim routes requests to configured, approved capacity.
Stay in control.

Check the provider and residency records for your requests. Change providers or models behind the same API, with model behaviour, latency and capacity tested for your application.

Rules you can inspect

Your account’s residency policy filters eligible routes before execution. A Sweden-only workload stays on routes configured for Swedish execution, including during failover. If no Swedish route is available, the request fails instead of running elsewhere.

When a provider fails or rejects a request because of load, Eldrim can try another configured route that meets the same policy. That requires another eligible route for the model. A stream can switch providers only before its first chunk.

Responses from a provider include that provider and its configured execution jurisdiction in response headers, and the same information is recorded with usage. You can check those records against your agreed policy. The records give traceability of the route; a geographic commitment also rests on the provider’s agreement and our operational controls. Independent hardware attestation is not available today.

Data handling, logs and any additional security or compliance requirements are reviewed as part of onboarding. How Eldrim approaches control of the stack.

Start building

Read the developer guide for API examples, coding tools and workload scheduling.

Already know what you need? Request early access and tell us which models and residency policy you want. If you already have an account, open the customer console.

Use your existing OpenAI-compatible client with the Eldrim base URL and key. List the models available to your account, then choose one for your first request.

curl https://api.eldrim.net/v1/models \
  -H "Authorization: Bearer $ELDRIM_API_KEY"

# Use an id returned above as ELDRIM_MODEL.
curl https://api.eldrim.net/v1/chat/completions \
  -H "Authorization: Bearer $ELDRIM_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"model\": \"$ELDRIM_MODEL\",
       \"messages\": [{\"role\": \"user\", \"content\": \"Hej\"}]}"

# Response headers include:
# x-eldrim-provider: ...
# x-eldrim-residency: se

Capacity that fits your workload

We recommend an arrangement based on your volume, performance needs and residency rules.

Arrangement	When it fits
Pay as you go	Start testing with early-access models and pricing per million tokens.
Reserved capacity	Discuss predictable throughput and pricing for sustained workloads. Availability and terms need agreement.
Private Swedish capacity	Explore a dedicated setup at a named Swedish site as a design partner. This is a future offering.

Build on infrastructure without becoming dependent on it

Today Eldrim’s API is in early access, behind a common control layer. We test it on our own hardware; production capacity will come from Swedish data-centre providers, which we are arranging now. You choose the rules; we handle the routing and help you choose a workable capacity arrangement.

As we add capacity, our aim is to keep the interface to your application stable. Reserved capacity and future Eldrim-operated Swedish infrastructure would sit behind the same API.

Where we are going

Over time we intend to operate modular infrastructure, connect its heat to local energy systems, reduce consumptive cooling-water use, and design the computer around memory bandwidth, networking, energy and heat recovery.

Haparanda is a possible future site under investigation. Customer demand comes first; feasibility work on power, connectivity and heat recovery follows. These are engineering goals, with each step intended to fund and validate the next.