Any open model, one endpoint.

One OpenAI-compatible endpoint for the whole open-weight ecosystem, from an EU company, with your prompts never stored.

No sign-up form. A short email and we take it from there.

# change two lines, keep the rest of your code
client = OpenAI(
    base_url="https://api.aiqre.com/v1",
    api_key="aiq-...",
)

resp = client.chat.completions.create(
    model="your-model",
    messages=[{"role": "user", "content": "Hi"}],
    stream=True,
)

Why Aiqre

Built for teams that care where their data goes.

An EU company

Operated from the European Union under EU law, with a GDPR data processing agreement available before you send us anything.

Nothing is stored

Prompts and completions exist in memory for the length of the request and are then gone. We never train on your traffic.

Drop-in compatible

Standard OpenAI chat completions with streaming. Change the base URL and the model name; the rest of your code stays as it is.

Fast and consistent

Capacity is reserved for your workload, so response times stay steady instead of drifting when traffic rises.

Weights you can verify

Every model maps to a published open-weight release, so you always know exactly what answered your request.

Paperwork ready

A GDPR data processing agreement is available before you send a single request, not after a procurement fight.

Models

Any open model, deployed within days.

We are not limited to a fixed menu. If a model has open weights, we can serve it — including recent releases, and models the larger providers have not bothered to add.

New releases, quickly

We deploy open-weight releases while they are still current, not months later.

Up to 200B parameters

Dense and mixture-of-experts, quantised or full precision.

One endpoint for all of them

Switching model is a string change, never an integration.

Text, vision, code, speech and moderation models all run behind the same API. Already know what you want? Send the name and your expected volume, and we will confirm pricing and have it live shortly after.

How it works

Three steps, no forms.

Name the model

One email with the model you want, roughly what volume, and whether you need a data processing agreement.

We deploy it

Usually live within a few days. We confirm the price up front and send you an API key.

Point your code at us

Change the base URL and the model name. Everything else in your integration stays exactly as it is.

Pricing

Pay per token, or commit and pay less.

Every quote is based on the model and volume you tell us about. No tiers to guess between.

Pay as you go

Metered per token, billed monthly

Priced per million input and output tokens
Quoted per model once we know your volume
No minimum and no commitment
Rate limits sized to your traffic

Get a quote

Committed

Monthly volume commitment, lower rate

Better price per token in exchange for commitment
Throughput guaranteed for your workload
Priority during busy periods
Private models and a DPA on request

Talk to us

API

Everything you already know, at a different URL.

curl https://api.aiqre.com/v1/chat/completions \
  -H "Authorization: Bearer $AIQRE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "your-model",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

Streaming over server-sent events, token usage returned on every response, and the usual sampling controls: temperature, top_p, top_k, stop, seed and max_tokens.

Read the docs

Questions

The things people ask first.

Which models can you actually run?

Anything with open weights, up to roughly 200B parameters — dense or mixture-of-experts, quantised or at full precision. That covers effectively the entire open-weight ecosystem.

If you name something we have not run before, we will tell you honestly whether it fits and how long it will take.

How quickly can you add a model I need?

Usually within a few days. Weights download in minutes and the endpoint is a configuration change, not a deployment project.

Very large models, or ones needing an unusual runtime, can take longer. We will say so before you plan around it.

What happens to my prompts?

Nothing is written to storage. Prompts and completions exist in memory for the length of the request and are gone when it finishes.

We record token counts, timestamps and latency for billing. We never record content, and we never train on your traffic.

Where does inference run?

We are an EU company and can discuss data residency requirements before you onboard. Tell us what your compliance team needs and we will confirm in writing what we can commit to.

Can I run a private or fine-tuned model?

Yes. Send us the weights and we will host them for your use only. They are not added to any public catalogue and no other customer can reach them.

What if I want to leave?

You change the base URL back. There is no proprietary format, no lock-in, and no contract you have to unwind on the pay-as-you-go plan.

We delete your account data on request, and everything we held was billing metadata to begin with.

Do you have an uptime guarantee?

Committed customers get an availability target written into the agreement, along with what happens if we miss it.

On pay-as-you-go we publish what we actually achieve rather than promising a number we have not yet earned.