Change one line.

Set the base URL and the key. Code built on the OpenAI SDKs, LangChain, Cursor or OpenCode keeps working.

Dedicated GPU

A GPU that serves only you.

A virtual machine with its own NVIDIA GPU in a Swiss data centre, running vLLM for your account alone.

Private by design

Prompts and answers are not stored.

The logs keep token counts, not content. Nothing is used to train a model.

Pricing

A monthly price. No limit on tokens.

Solo starts at CHF 399 a month, with chat and API keys included. Cancel for the end of any month.

Plans

Pick the GPU. Keep the weights.

Apache 2.0 models from Qwen and Google. If you leave, they run on your own hardware. Your fine-tune runs too, if it fits.

Agent · CHF 19 a month

An agent that keeps working.

Write to it where you already are. It runs tasks on a schedule, uses your tools, and remembers.

app.py

from openai import OpenAI
 
client = OpenAI(
    base_url="",
    api_key="",
)
 
r = client.chat.completions.create(
    model="",
    messages=[{"role": "user",
               "content": "Summarise this…"}],
    stream=True,
)

200 OK · api.llmaas.ch · streaming

Data centre in Switzerland

Your appOpenAI SDK

LLMaaS gatewayapi.llmaas.ch

Your GPUNVIDIA L4 · 24 GBGemma 4 26B-A4B

Serves your account, nobody else

Streaming, tools, images

Speed depends on your load alone

stream

# gateway log

prompt: Summarise this report…

answer: The report covers three…

tokens_in=412 tokens_out=186

prompt

Billed per tokenLLMaaS Solo, flat

Tokens this month0

Your invoiceCHF 399

Agentalways on

Encrypted memory, yours to download

Scheduled tasks, every morning or hour

Shell, files, git and MCP tools

LLMaaS

Your own LLM endpoint.
Hosted in Switzerland.

Open-weight modelsOpenAI-compatible APIUnlimited tokensOperated from Prilly, VD

llmaas.ch