# Berget AI

> Berget AI is a Swedish AI inference platform offering inference and AI infrastructure for private and public sector organisations. Berget is incorporated in Sweden with 100% Swedish ownership — data never leaves Sweden and is fully protected from US Cloud Act and FISA 702.

## Use Cases

- **GDPR-compliant chatbot:** Deploy a customer-facing assistant where all data stays in Sweden — suitable for HR, legal, and healthcare applications.
- **EU-sovereign RAG pipeline:** Combine Berget's embeddings, reranking, and chat endpoints to build retrieval-augmented generation entirely within Swedish infrastructure.
- **OpenAI migration:** Replace OpenAI with one line of code for teams that need EU data residency without rewriting existing applications.
- **Swedish speech-to-text:** Transcribe Swedish-language audio with KB-Whisper-Large, optimized for Swedish by the National Library of Sweden.
- **Document intelligence:** Extract structured data from PDFs, Word files, and images — all processed in Sweden.

## Getting Started

[Create an account](https://console.berget.ai/auth) and then create API key:

```bash
npx berget auth login
npx berget api-keys create --name my-key
```

Or manage keys at `https://console.berget.ai`.

Use OpenAI-compatible tools and SDKs by setting base URL to `https://api.berget.ai/v1`:

```python
from openai import OpenAI

client = OpenAI(
    base_url="https://api.berget.ai/v1",
    api_key="your-berget-api-key"
)
```

## Compliance

- **GDPR**: Compliant. Data processed and stored exclusively in Sweden.
- **US Cloud Act / FISA 702**: Protected — 100% Swedish ownership, no US parent company, no US jurisdiction.
- **EU AI Act**: Compliant.
- **DORA**: Suitable for financial sector use.
- **NIS2**: Suitable for critical infrastructure operators.

## Plans

| Plan       | Price      | Wallet Refill                   | Notes                                                                        |
| ---------- | ---------- | ------------------------------- | ---------------------------------------------------------------------------- |
| Free       | €0/month   | €5 signup bonus (one-time)      | Pay as you go, rate limited, community support                               |
| Starter    | €25/month  | €25/month                       | Automatic billing, email support, cancel any time                            |
| Team       | €50/month  | €55/month (+10% extra credits)  | Priority support                                                             |
| Enterprise | €500/month | €550/month (+10% extra credits) | No rate limits, dedicated support, invoice + Peppol (EU e-invoicing) billing |

Usage above the included amount is charged on top. Unused balance rolls over monthly. All plans support credit card pre-payment.

## Models

For full model catalogue and pricing, see [Models](https://console.berget.ai/models).

List available models via API: `GET /v1/models` (public, no auth required).

To test and evaluate model capabilities, use the [Testmatris](https://berget-models.lovable.app/) — Berget's official model evaluation tool.

### Chat and Coding

| Model                  | ID                                              | Price (EUR/M tokens in/out) |
| ---------------------- | ----------------------------------------------- | --------------------------- |
| Llama 3.1 8B Instruct  | `meta-llama/Llama-3.1-8B-Instruct`              | €0.20 / €0.20               |
| Llama 3.3 70B Instruct | `meta-llama/Llama-3.3-70B-Instruct`             | €0.90 / €0.90               |
| Mistral Small 3.2 24B  | `mistralai/Mistral-Small-3.2-24B-Instruct-2506` | €0.30 / €0.30               |
| GPT-OSS-120B           | `openai/gpt-oss-120b`                           | €0.30 / €0.90               |
| GLM 4.7 (358B MoE)     | `zai-org/GLM-4.7`                               | €0.70 / €2.30               |

### Embedding

| Model                          | ID                                        | Price   |
| ------------------------------ | ----------------------------------------- | ------- |
| Multilingual-E5-large-instruct | `intfloat/multilingual-e5-large-instruct` | €0.03/M |
| Multilingual-E5-large          | `intfloat/multilingual-e5-large`          | €0.03/M |

### Reranking

| Model           | ID                        | Price   |
| --------------- | ------------------------- | ------- |
| bge-reranker-v2 | `BAAI/bge-reranker-v2-m3` | €0.10/M |

### Speech-to-Text

| Model            | ID                       | Price                |
| ---------------- | ------------------------ | -------------------- |
| KB-Whisper-Large | `KBLab/kb-whisper-large` | €3.00 / 1000 minutes |

### OCR

| Model        | ID                         | Price              |
| ------------ | -------------------------- | ------------------ |
| DeepSeek-OCR | `deepseek-ai/DeepSeek-OCR` | €0.40/M OCR tokens |

## API Endpoints

All endpoints at `https://api.berget.ai/v1`. Auth: `Authorization: Bearer <api-key>`.

- `POST /v1/chat/completions` — OpenAI-compatible chat. Supports streaming, tool/function calling, vision (base64 or URL), JSON mode, structured output.
- `POST /v1/embeddings` — Vector embeddings.
- `POST /v1/rerank` — Rerank documents by relevance.
- `POST /v1/audio/transcriptions` — Speech-to-text, Swedish-optimised. Max 100 MB. Formats: json, srt, vtt.
- `POST /v1/ocr` — Extract text from PDF, DOCX, PPTX, HTML, images. Output: Markdown or JSON. Supports async for large documents.
- `GET /v1/models` — List available models with capabilities and pricing (public).

For full API documentation, see [Berget API docs](https://api.berget.ai).

For current status of our services, see [status page](https://status.berget.ai).

### Chat Example

```bash
curl https://api.berget.ai/v1/chat/completions \
  -H 'Authorization: Bearer BERGET_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "openai/gpt-oss-120b",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello!"}
    ]
  }'
```

### Embeddings Example

```bash
curl https://api.berget.ai/v1/embeddings \
  -H 'Authorization: Bearer BERGET_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "intfloat/multilingual-e5-large-instruct",
    "input": "This is a sample text to create embeddings for."
  }'
```

### Rerank Example

```bash
curl https://api.berget.ai/v1/rerank \
  -H 'Authorization: Bearer BERGET_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "BAAI/bge-reranker-v2-m3",
    "query": "What is machine learning?",
    "documents": [
      "Machine learning is a branch of artificial intelligence.",
      "Natural language processing is used in many applications.",
      "Deep learning is a subset of machine learning."
    ],
    "top_n": 2,
    "return_documents": true
  }'
```

### Speech-to-text Example

```bash
# Upload a file
curl https://api.berget.ai/v1/audio/transcriptions \
  -H 'Authorization: Bearer BERGET_API_KEY' \
  -F 'file=@audio.mp3' \
  -F 'model=KBLab/kb-whisper-large'

# Or pass a URL
curl https://api.berget.ai/v1/audio/transcriptions \
  -H 'Authorization: Bearer BERGET_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{"file_url": "https://example.com/audio.mp3"}'
```

### OCR Example

```bash
curl https://api.berget.ai/v1/ocr \
  -H 'Authorization: Bearer BERGET_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "deepseek-ai/DeepSeek-OCR",
    "document": {"url": "https://example.com/document.pdf", "type": "document"},
    "options": {"outputFormat": "md"}
  }'
```

For large documents, add `"async": true` — the response returns a job ID. Poll `GET /v1/ocr/{job_id}` for the result.

## Questions and Support

- [Community forum](https://umami.berget.ai/q/nnlADzkKU)
- [Support](mailto:support@berget.ai)
- [Sales / DPA enquiries](mailto:andreas@berget.ai)
- [Privacy](mailto:privacy@berget.ai)