Skip to content
deeplit logo
Start on deeplit® Cloud
[ PRICING ]

Know the bill before you run anything.

deeplit® Cloud is priced by the GPU hour, however much you use it. deeplit® Private runs on your own hardware, priced to your deployment.

  • deeplit® Cloud

    Recommended

    Start today. Pay for the hours you run.

    Pay as you go

    from €1 / GPU hour

    • No platform fee
    • Model serving and full catalog
    • Users, roles, and API keys
    • Audit log, SSL, monitoring
    • Billing stops when you stop it
    • Same day start
    Start on deeplit® Cloud

  • deeplit® Private

    Your own hardware, licensed as one platform.

    Contact us

    Priced on your users, load, and hardware.

    • Your own server or data center
    • Nothing leaves your perimeter
    • Fully disconnected if you need it
    • The same core, licensed with your modules
    • The deployment stays yours
    • About two weeks to start
    Talk to us

GPU classes run from €1 an hour. Full list in the console.

[ MODULES ]

Add what the workload needs.

The core runs models, users, keys, audit, SSL, and monitoring. Modules sit on top, licensed on Cloud as you need them and alongside the platform on Private.

Ask what a module costs

  • Agentic RAG

    Searches your documents in steps and answers from what it finds.

  • GraphRAG

    Answers from how the facts in your documents connect to each other.

  • NLQ

    Ask your databases a question in plain words and get the rows back.

  • Fine tune

    Trains a model on your own examples, inside your perimeter.

[ PRICING FAQ ]

Questions procurement asks.

There is no platform fee on deeplit® Cloud. You pay for the GPU hours you run, and the core is what runs on them: model serving, the catalog, users, API keys, audit log, SSL, and monitoring. Modules are licensed separately.

Buying GPUs and self-hosting runs about €170k to €390k in year one, where hardware is a fifth of it and engineers are the rest. Building a private deployment in your customer's cloud is about €60k to €120k of engineering. Letting the deal slip costs about €50k to €250k, and the next buyer asks the same question.

A GPU does bounded work in an hour, so cost has a ceiling a token meter does not. Heavy users cost what light users cost, and finance gets one number to forecast. Your dashboard counts tokens; your invoice never does.

Private is priced on your users, your load, and your hardware, and every deployment is its own case. A short call settles all three, with the modules you license. Cloud needs none of that, which is why you can start it yourself.

You provide a GPU server you own, in your building or colocation, plus deployment access. Everything above that is ours to build and hand over, and running it needs no ML expertise on your side.

When traffic is spiky or occasional. Flat wins on predictability and sustained use, so light bursty work can be cheaper on a public AI API. We will say so.

No. deeplit® Cloud is funded up front, so you only run what you have paid for. An instance bills while it runs and stops when you stop it.

[ NEXT STEP ]

Start on the core. Add modules when you need them.

No platform fee on Cloud, and nothing to commit to. For your own hardware, talk to us.

Start on deeplit® CloudTalk to us about Private