Simple hourly pricing for your own deployment

Every instance runs one model on a dedicated GPU. You pay per started hour while it runs, and nothing when it does not. No tokens, no seats, no minimum commitment.

One price per model class

Pick the class that fits the work. The price covers the model and its dedicated hardware.

Entry

Compact, efficient models

from €4.99 / hour

excl. VAT · €5.94 incl. 19% VAT

  • Muse Glimmer 30B
  • Gemma 4 31B
  • Qwen3.8 27B

Red Team: Verified email and intended use

Pro

For demanding workloads

from €19.99 / hour

excl. VAT · €23.79 incl. 19% VAT

  • GLM 5.3 Flash
  • Qwen3.8 Flash Next
  • MiMo V2.6 Flash RL

Red Team: Full verification required

Frontier

Frontier-scale models

from €29.99 / hour

excl. VAT · €35.69 incl. 19% VAT

  • GLM 5.3
  • DeepSeek V4.1 Flash 1017
  • Kimi K3

Red Team: Full verification, business customers only

Prices exclude VAT. Consumers pay the VAT of their country, shown before the deploy.

How you pay

Two ways to pay, depending on the model class. Payments are processed by Stripe.

Entry and Basic

Pay as you go

Start an instance without an advance payment and pay for the hours it actually ran.

  • No upfront payment
  • Charged at the end, at month end, or once €250 are open
  • Stop at any time
Pro and Frontier

Reserved up front, billed on actual use

You choose a planned running time. That amount is reserved on your payment method before the deploy, renewed automatically if the instance runs longer, and only the actual usage is billed.

  • Reservation shown at the deploy
  • Only actual usage is charged
  • Reservation released afterwards

Card and payment details are handled by Stripe. We do not store complete card data.

Included in every instance

The same foundation under every class, from Entry to Frontier.

Dedicated GPU

Your instance runs on its own GPU. Nothing is shared with other customers.

EU hosting

Hosted only with EU GPU providers certified to ISO/IEC 27001.

OpenAI-compatible API

A private endpoint that works with the tools and agents you already use.

No prompt storage

Prompts and outputs are never stored, and customer data is not used for training.

No token metering

A flat hourly price. No per-token fees and no shared rate limits.

Blue and Red Team

Both variants of a class cost the same, where the model licence allows a Red Team variant.

Estimate your cost

Choose a class and how long the instance runs. You only pay for the hours it is up.

Model class
Basic

€399.60

40 h × €9.99 per hour

Indicative, based on the starting price of the class, excl. VAT. You pay only while an instance runs, and every started hour counts as a full hour.

How you are billed

The clock starts when your endpoint is healthy, and stops when you delete the instance.

  1. DeployInstance starts and loads the modelNot billed
  2. HealthyFirst successful health check. The clock starts.Billed
  3. RunningEvery started hour counts as a full hourBilled
  4. StopThe instance is deleted and the remaining hours are chargedNot billed

Pricing questions

When does billing start?

At the first successful health check. Time spent starting up, including failed deploys, is not billed.

How do I pay?

Entry and Basic are pay as you go: hours are charged when the instance ends, at the latest at month end, and whenever €250 are open. For Pro and Frontier you choose a planned running time, that amount is reserved, and only the actual usage is charged. Payments are processed by Stripe.

Do I pay per token?

No. The price is per hour on a dedicated GPU, so heavy use costs the same as light use while the instance runs.

What happens when I stop an instance?

Stopping deletes the instance and closes the billable interval. The hours not yet charged are charged then.

Does the Red Team variant cost more?

No. Blue Team and Red Team variants of a model class have the same price. Some Red Team models need verification before you can deploy them.

Do I pay when an instance is down?

No. Hours in which an instance is not healthy because of a problem on our side or at the GPU provider are not billed.

Can I cap my spending?

Yes. You can set a monthly spending limit, get a warning at 80 per cent, and choose to stop your instances when the limit is reached.

Are prices with or without VAT?

Prices are shown without VAT. Business customers in Germany pay 19% VAT on top, business customers in other EU countries with a valid VAT ID are invoiced under the reverse charge procedure. Consumers pay the price including the VAT of their country, shown before the deploy. For consumers in Germany this is the price shown plus 19%.

Are these final prices?

They are indicative starting prices. Hardware and prices per model stay provisional until our launch benchmarks are complete.

Request early access

Tell us about your team and the models you want to run. We will contact you as soon as your tier is available.