Support and salesSupport / sales
Unlimited Basic and ProFrom €14.95/month · Unlimited models subject to fair use

Deployments on QDivZero

Deploy models. Not infrastructure.

Pay for active time.

Deploy Hugging Face models or your own on dedicated GPU infrastructure. QDivZero automatically finds the lowest-cost compatible option in your chosen region and prepares the inference environment for you.

No server management. No per-token fees. Billed per second of active compute.

01 / Understand compute

More inference.No more per-token costs.

Make the most of your GPU capacity, not a token allowance.

With deployments on QDivZero, your cost depends on the GPU resources you use and the time they remain active, rather than the number of tokens you process. Run as many requests as your deployment’s capacity allows, with no additional per-token charges.

Check the per-second price before deploying and decide when to keep your infrastructure running. An especially useful alternative for applications with intensive or continuous AI usage.

01 / Dedicated capacity

Your GPU. Its full capacity.

No per-token fees.

Run your model on dedicated GPU resources and use their capacity to process requests without paying individually for generated tokens.

Performance depends on the model, configuration, and allocated infrastructure.

02 / Flexible usage

Start when you need it. Pay while it’s active.

Per-second billing.

Check your deployment’s price before starting it and control when it remains running.

Schedule active hours or use serverless to adapt compute time to your demand.

Active time is billed even when the model receives no requests. Shutdown and serverless options depend on the deployment configuration.

02 / The complete deployment, without managing infrastructure

You choose the model. QDivZero handles the rest.

Everything behind a deployment, taken care of.

Running a model in production involves much more than finding a GPU. You need to prepare its environment, configure inference, connect it to your applications, and manage availability.

With QDivZero Compute, you can do it all in one workflow, without setting up or maintaining each component separately.

01 / Configure

Your model. Your way.

Select Hugging Face, private, or fine-tuned models. Adjust inference settings and use persistent storage to retain and share data between runs.

Custom models · Serving · Persistent disks

02 / Deploy

The right GPU. Without searching for it.

QDivZero checks your model’s requirements and automatically selects the lowest-cost compatible infrastructure available in your chosen region across its multi-provider network.

Multi-provider scheduler · Dedicated GPU · Regions

03 / Connect

From your model to your application.

Access your models through OpenAI-compatible endpoints. Integrate private inference with HTTPS/TLS-encrypted connections and zero content logs by default.

OpenAI-compatible API · Endpoints · Security

04 / Optimize

Switch on when you need it.

Schedule starts and stops with cron or use serverless to activate models on demand. Monitor GPU, CPU, RAM, token, and performance metrics to optimize each deployment.

Cron · Serverless · Observability

Want to know where your models run or work with us?

Meet our providers

03 / Intelligent infrastructure selection

Don’t search for providers. Find capacity.

A multi-provider network that works for you.

Not every GPU can run every model, and providers differ in capacity and pricing.

QDivZero analyzes your deployment requirements and automatically finds the lowest-cost compatible option available across its infrastructure providers, within your selected region.

You choose the model and location. Compute finds where to run it.

How automatic selection works

  1. Requirements

    Model and workload

  2. Region

    Selected location

  3. Selection

    Lowest-cost compatible capacity

04 / Compute adapts to your workload

Your compute doesn’t have to stay on all the time.

Control when it runs. Optimize what you consume.

A production application doesn’t always need the same capacity. With QDivZero, you can keep models available, schedule runtime hours, or use serverless to start them on demand and stop them after a configurable period of inactivity.

Monitor deployment performance and adjust settings to match your workload.

Pay for active compute time and avoid keeping resources running when you don’t need them.

Scheduled execution

Define start and stop schedules using cron rules.

Serverless

Start models on demand and stop them after periods of inactivity.

Observability

Monitor GPU, CPU, RAM, tokens, and performance to optimize your resources.

05 / FAQ

Frequently asked questions about model deployments

Answers to common questions about dedicated GPU infrastructure, billing for active time, provider selection, and deploying custom or Hugging Face models.

How is a deployment billed on QDivZero?

The cost is calculated per second based on the GPU resources allocated and how long they remain active. There are no additional charges for tokens processed during inference by the deployed model.

Do I pay even if my model receives no requests?

Yes. Billing continues while the infrastructure remains active, even when there are no requests. You can schedule your deployments to start and stop or use serverless to adjust active time.

How does QDivZero select GPU infrastructure?

QDivZero analyzes the model requirements and automatically finds the lowest-cost compatible option available across its providers within your selected region.

Can I deploy private or fine-tuned models?

Yes. You can deploy compatible models from the Hugging Face catalog, private models, or fine-tuned models, provided they meet the technical requirements for execution.

What is the difference between public models and dedicated deployments?

Public models are ready to use through the API and are billed by token. With dedicated deployments, you run a model on exclusive GPU resources and pay for the time they remain active, regardless of the tokens processed.

Choose your model. Put compute to work.

Select your model, choose a region, and let QDivZero find the right GPU infrastructure. Check the cost per second before deploying.