Skip to content

Infinized LLMEarly access

Language models,
run in the Netherlands.

Infinized LLM serves open-weight Qwen3 models from our own servers in the Netherlands. Call it through an OpenAI-compatible API, or route your AI agents and flows to it from Infinized.

Sign in with InfinizedRead the docs
Chat Completions · example
POST https://api.openai.com/v1/chat/completionsAuthorization: Bearer $OPENAI_API_KEYContent-Type: application/json{  "model": "gpt-5.4-mini",  "stream": true,  "messages": [    { "role": "user", "content": "Summarise in one sentence: Hi, I was charged twice for order 4471 this morning. Could you refund one of the payments?" }  ]}

Response

Switch to Infinized LLM: only the highlighted parts change, and the answer streams in here.

Already on Chat Completions? Change the base URL and the key, pick an Infinized model, and keep your code. Streaming and non-streaming both work.

How it works

Three steps to your first request.

Then point your OpenAI-compatible client at llm.infinized.com/v1.

  1. Sign in with Infinized

    Use the Infinized account you already have. There is no separate password to keep.

  2. Add a payment method

    Add a billing profile and a card or SEPA Direct Debit. Pay afterwards for the tokens you use, as the model runtime counts them, invoiced automatically as usage grows. No subscription needed.

  3. Create a key

    The secret is shown once. Give a key a monthly spending limit, and revoke it whenever you like.

Model routing

Route your AI agents to it.

Link a key to your Infinized workspace in one click. Infinized LLM then sits beside OpenAI, Anthropic and Mistral as a model provider, and you choose which AI agents and flows run on models Infinized runs itself.

Chosen per AI agent and flow
Move the simple, high-volume steps. The rest keep the model they have.
Recorded on every run
Each run records the model it ran on and why, as it does for every provider.
How routing works
Your workspace · model routingExample
  • OpenAI
  • Anthropic
  • Mistral
  • Infinized LLM
  • Ticket triageFlowSorts every new ticket into a queueinfinized-nano-v1Netherlands
  • Order e-mailsFlowPulls the order number and the question out of each e-mailinfinized-mini-v1Netherlands
  • SanneAI agentDrafts replies to complex complaintsclaude-sonnet-5Own key

2 of 3 routed to Infinized LLM. Every run still records the model it ran on, and why.

The models

Small models for well-defined work.

Both are built on open-weight Qwen3 models (Apache 2.0) and served by Infinized with llama.cpp on CPU servers in the Netherlands. They read and write Dutch and English, and suit high-volume language tasks that should stay in the Netherlands.

  • infinized-nano-v1

    Based on Qwen3 0.6B · Apache 2.0

    Context
    2,048
    tokens
    Output
    512
    tokens at most
    Languages
    NL · EN
    Dutch and English

    Good for

    • Classification
    • Simple extraction
    • Short rewrites

  • infinized-mini-v1

    Based on Qwen3 1.7B · Apache 2.0

    Context
    4,096
    tokens
    Output
    512
    tokens at most
    Languages
    NL · EN
    Dutch and English

    Good for

    • Short summaries
    • Extraction
    • Answers over context you supply

Privacy and where it runs

Your prompts are not stored.

Infinized LLM keeps only a request’s metadata: token counts, timing, status and price. That is what your bill is made of.

Run in the Netherlands
The models run on Infinized’s own servers in the Netherlands.
Prompts and answers not stored
What you send and what comes back is processed, not kept.
Not used for training
Your content is not used to train any model.
Security at Infinized

Start with one task.

Sign in with your Infinized account, add a payment method and create a key. Your first request is one base URL away.

Sign in with InfinizedRead the docs