Enterprise
Your own deployment in Europe, a model version that stays fixed, contracts your purchasing can sign, and people who help you go live.
New to Strom? What it does, how fast and at what price is on the home page.
Once Strom decides real cases, more people get a say than the team that built the integration. Here is what they need, and what we do about it.
An update that shifts behaviour shifts every automated decision with it.
What we do: Your deployment keeps its model version. We announce updates in advance and test the new version on your examples before you switch.
Auditors and your own team will ask why a case was decided the way it was.
What we do: Every answer comes with its probability and is logged with the request. You set the threshold, and we agree with you how long the log is kept.
Your security team and data protection officer review every new processor.
What we do: Processing in Europe, a data processing agreement, a public subprocessor list, and answers to your security questionnaire.
Prepaid credit on a card rarely fits how companies buy.
What we do: A contract with a monthly invoice, priced for the capacity you need.
A checkout or a camera feed can’t queue behind other customers’ traffic.
What we do: GPUs reserved for you, sized to your volume, with uptime and response times in an SLA.
Purchasing will ask what happens if a young vendor changes course.
What we do: We export your data on request. The API takes the same request format as TypeSafe’s Jev, so switching is a configuration change, not a rewrite. And the contract fixes how far in advance we would announce an end of service.
Both run the same Strom model behind the same API. Dedicated adds your own capacity, your region and guarantees.
| Public API | Dedicated | |
|---|---|---|
| Hosting | Shared GPUs in Paris | GPUs reserved for you, in France, Germany or another European region |
| Throughput | Up to 1,200 requests a minute, 16 at a time | Sized to your volume |
| Under load | Requests wait when the shared GPUs are busy | Only your traffic, so latency stays predictable |
| Availability | No SLA during the beta | Uptime agreed in an SLA |
| Support | Documentation and email | A named engineer and a shared channel |
| Model updates | Updated for all customers at once | Your version stays until you choose to update |
| Data retention | Request contents deleted after 90 days | Agreed in your contract |
| Billing | Prepaid credit, $0.042 per million input tokens | A contract and a monthly invoice for your capacity |
In both cases the model stays with us. There’s no self-hosted or on-premise version. You call Strom over the API, and we run the GPUs.
The team that trains Strom helps you put it to work.
One person on our side who knows your use case, from the first call to production.
A shared Slack or Teams channel for quick questions, alongside email.
Together we turn your decisions into questions and options that Strom answers well.
Before you go live, we test Strom on your own examples and show you where it’s right and where it isn’t.
How fast we respond, by severity, is part of the SLA.
New domains or languages, unusual input sizes or integrations. Tell us what you need, and we’ll work out how.
Decisions often run on personal data, and photos, scans and camera frames are often the most sensitive part. Strom keeps the path short, so your data protection officer gets a simple answer.
In the form below. We get in touch to hear which decisions you want to automate.
In a call with an engineer. The deployment, the region, the service level agreement and anything your use case needs beyond that.
After an evaluation on your own examples. Your requests use the same API and request format as the public endpoint.
Leave your work email and we’ll get in touch. Prefer to write first? sales@uprelic.com.
