European AI inference.
Your data stays in Europe.
beServe proxies OpenAI-compatible requests to inference hosted in Paris. Change two lines of code, keep everything else.
No token metering surprises. Flat monthly plans, request-capped.
# your code today
from openai import OpenAI
client = OpenAI(
api_key="sk-…"
)
# with beServe — change two lines
client = OpenAI(
api_key="bs_live_…",
base_url="https://api.beserve.eu/v1"
)
response = client.chat.completions.create(
model="deepseek-v4-flash-0731",
messages=[{"role": "user",
"content": "Bonjour !"}]
)Every request makes a round trip inside the European Union. No fourth stop.
Data stays in Europe
Inference runs in Scaleway’s Paris datacenter. Prompts, responses, logs — nothing crosses an EU border.
Two lines to migrate
You keep your OpenAI SDK, your prompts, your error handling. Change the API key and the base URL, done.
Euro billing, TVA included
European billing, EUR pricing, proper invoices. B2B with a TVA number? Reverse charge, 0%.
Streaming that streams
Server-sent events flow chunk by chunk. First token in hundreds of milliseconds, not seconds.
Model catalog
Real prices, in euros, per 1K tokens HTVA. Streaming included on every plan.
| Model | Context | Input / 1K | Output / 1K |
|---|---|---|---|
| deepseek-v4-flash-0731 | 128K | €0.0006 | €0.0012 |
| gpt-oss-120b | 128K | €0.0002 | €0.0008 |
| gemma-4-26b-a4b-it | 128K | €0.00035 | €0.0007 |
| mistral-medium-3.5-128b | 128K | €0.0021 | €0.0105 |
| pixtral-12b-2409 | 128K | €0.0003 | €0.0003 |
| glm-5.2 | 128K | €0.0025 | €0.0077 |
Scroll the table sideways for output prices →
Ready in three steps
- 1 Create an account
Pick a plan, verify your email.
- 2 Generate a key
From the dashboard, shown once, starting with
bs_live_. - 3 Call the API
Same OpenAI SDK. New key and base URL. That's the whole migration.