# Kurrens documentation

> Fast, private inference for open models through one OpenAI-compatible API.

Kurrens is an inference provider for open-weight models. We run models such as DeepSeek, Kimi, GLM, Qwen, and MiniMax on GPUs we operate in Malaysia, Indonesia, and Ireland,
and serve them through one **OpenAI-compatible API**, with **zero data retention**: prompts and completions are processed
in memory and never stored or used for training.

<PreLaunch />

## What you can do

<CardGrid>
  <Card title="Serverless inference">
    Call open models per token with the OpenAI SDK you already use. Streaming, tool calling, structured outputs, and prompt
    caching where the model supports them.
  </Card>
  <Card title="Dedicated endpoints">
    Reserve GPUs for a single model and get a private base URL with consistent latency and custom rate limits.
  </Card>
  <Card title="Zero data retention">
    Inputs and outputs live in memory for the length of the request. We keep only request metadata for billing.
  </Card>
  <Card title="Declared precision">
    Every model lists its Hugging Face weights, quantization, context window, and price.
  </Card>
</CardGrid>

## Start here

<CardGrid>
  <LinkCard title="Quickstart" description="Make your first request with the OpenAI SDK or cURL." href="/docs/getting-started/quickstart" />
  <LinkCard title="Model catalog" description="Search every model we serve, with context length and precision." href="/docs/getting-started/models" />
  <LinkCard title="Zero data retention" description="Exactly what we store — and what we don't." href="/docs/data-security/zero-data-retention" />
  <LinkCard title="Regions" description="Johor, Jakarta, and Dublin — where your requests run." href="/docs/getting-started/regions" />
  <LinkCard title="Precision & versioning" description="The same weights behind an id, always." href="/docs/models/precision-and-versioning" />
  <LinkCard title="FAQ" description="Access, models, data, limits, and billing." href="/docs/help/faq" />
</CardGrid>

## Base URL

```text
https://api.kurrens.ai/v1
```

Any client that accepts an OpenAI base URL works with Kurrens.