BreignHUB
Sovereign inference · Hosted by DEEP · Luxembourg

Every open model. One sovereign endpoint.

BreignHUB gives your applications a single OpenAI-compatible API to the open-weight models DEEP runs in its Tier IV data centres in Luxembourg. Your prompts stay in the country. Your data is never used for training.

Get an API keySee what you get
7
models online
100 %
inference in Luxembourg
0
prompts retained
chat_completions.sh
curl https://hub.breign.eu/api/v1/chat/completions \
  -H "Authorization: Bearer $BREIGNHUB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mistral-large-3",
    "messages": [
      { "role": "user",
        "content": "Summarise this supplier contract." }
    ]
  }'
Your SDK. One base URL.200 OK

OpenAI SDK

Chat, models, embeddings

/v1

Chat · Embeddings · Audio

Infrastructure, security and compliance by DEEP · POST Luxembourg

  • Tier IV data centres
  • ISO 27001
  • ISO 27701
  • CSSF-ready
  • GDPR · AI Act

One endpoint

Everything your product asks a model to do

Chat, code, documents, retrieval and voice — served from the same authenticated base URL, billed on the same invoice, logged in the same audit trail.

  • Chat & reasoning

    Frontier open-weight models with extended reasoning, from 7B assistants to 400B+ mixture-of-experts.

  • Code & agents

    Long-context coding models with native tool calling and structured output for agent runtimes.

  • Vision & documents

    OCR, table extraction and page understanding for contracts, invoices and scanned archives.

  • Embeddings & reranking

    Multilingual vectors and cross-encoders for RAG pipelines that must stay inside the perimeter.

  • Speech & transcription

    Meeting and call transcription in French, German, English and Luxembourgish, with diarisation.

  • Dedicated capacity

    Reserve GPUs for a model that is yours alone — fine-tuned weights, fixed throughput, fixed price.

The catalogue

7 models running today

Open weights only — no closed API is proxied. Every model is deployed, patched and monitored by DEEP, so you can pin a version and keep it for as long as you need it.

Browse the full catalogue

Sovereignty

Your prompt is computed in Luxembourg. Full stop.

Not "EU region". Not "processed in Europe where possible". The GPUs are in DEEP's Tier IV data centres on Luxembourgish soil, operated by a company owned by the Luxembourg State. No request leaves the building.

Read the security briefTalk to an architect

  • No training on your data

    Prompts and completions are discarded at the end of the request. Zero retention is the default, not an enterprise upsell.

  • Encrypted end to end

    TLS 1.3 in transit, keys managed in DEEP's HSM. Private connectivity available over POST Telecom links.

  • Audit-ready by design

    Signed, immutable access logs you can hand to the CSSF or a data protection authority without a discovery project.

  • AI Act aligned

    Model cards, provenance and versioning published for every endpoint, so your own conformity file writes itself.

Migration

Point your OpenAI client at one more base URL

BreignHUB exposes a real OpenAI-compatible surface — chat completions with streaming, models and embeddings — and translates it to the Breign gateway behind your organization. Your SDK, LangChain chain or agent framework keeps working.

  • Streaming chat completions, model listing and embeddings, on the OpenAI wire format
  • Fields the gateway cannot honour are refused with a named error, never silently dropped
  • One key carries your organization; revoke it from the dashboard and it stops working
from openai import OpenAI

client = OpenAI(
    base_url="https://hub.breign.eu/api/v1",
    api_key=os.environ["BREIGNHUB_API_KEY"],
)

resp = client.chat.completions.create(
    model="mistral-large-3",
    messages=[
        {"role": "system", "content": "You are a compliance analyst."},
        {"role": "user", "content": "List the termination clauses."},
    ],
    stream=True,
)

Part of the BREIGN platform

Raw models here. Governance, agents and human oversight next door.

BreignHUB is the inference layer. When you need prompt filtering, approval workflows, RAG over your own repositories or supervised agents, the same account opens BREIGN — the sovereign agentic platform built on the same infrastructure.

BREIGN Studio

Build and supervise agents

BreignMate

The assistant your teams use daily

Discover BREIGN

Pricing

Per token, per GPU-hour, or per year

Invoiced in euros by a Luxembourgish entity. No egress fees, no minimum commitment on the pay-as-you-go tier.

  • Pay as you go

    Self-serve. Add a card, get a key, start streaming tokens in under two minutes.

    On request

    per 1M tokens

    Create an account

  • Business

    Most chosen

    Volume rates, per-project quotas, invoicing and a signed data processing agreement with DEEP.

    On request

    per month

    Contact sales

  • Dedicated

    Reserved GPUs, your fine-tuned weights, guaranteed throughput and private connectivity.

    On request

    per GPU-hour

    Talk to an architect

Ship your first sovereign completion this afternoon

Free credits on sign-up. No sales call required to start.

Get an API keyRead the docs