🇩🇪 EU-Native Inference

Your prompts
vanish.
The answers stay.

OpenAI-compatible AI inference API on bare-metal in Germany. No prompt logging. No data leaves the EU. GDPR compliant by design — not by policy.

Get API Key → See Models
api.cheshir.dev
# Drop-in replacement — change one line

from openai import OpenAI

client = OpenAI(
  api_key="ck-your-api-key",
  base_url="https://api.cheshir.dev/v1"
)

response = client.chat.completions.create(
  model="qwen2.5-72b",
  messages=[{"role": "user",
            "content": "Explain GDPR in 3 points"}]
)

▸ 200 OK · 1.18s · 0 bytes stored
🇩🇪
GermanyHetzner Datacenter
⚡
Bare MetalNo virtualisation
🔒
GDPR ReadyEU data residency
✓
OpenAI CompatibleDrop-in replacement
🙈
Zero LoggingPrompts never stored

Available Models

Three models for every workload

From fast responses to deep reasoning — choose the right model or use all three through the same API.

General

Qwen 2.5 72B

Best overall quality. Excellent at analysis, writing, coding, and multilingual tasks including Russian and Ukrainian.

Speed4–7 tok/s
Context8 192 tokens
QuantizationQ4_K_M
Model IDqwen2.5-72b
Reasoning

DeepSeek R1 32B

Chain-of-thought reasoning. Built for math, code, logic, and structured tasks that require step-by-step thinking.

Speed8–12 tok/s
Context8 192 tokens
QuantizationQ4_K_M
Model IDdeepseek-r1-32b
Fast

Llama 3.1 8B

High throughput, low cost. Ideal for classification, summarisation, and high-volume workloads where speed matters.

Speed25–40 tok/s
Context8 192 tokens
QuantizationQ8_0
Model IDllama-3.1-8b

Pricing

Simple, predictable pricing

No hidden fees. No data retention. Pay only for what you use.

Developer

$0.08

per 1M tokens · Llama 8B

Pay-as-you-go. Perfect for testing and low-volume workloads.

  • Llama 3.1 8B access
  • OpenAI-compatible API
  • No monthly minimum
  • Zero prompt logging
  • EU data residency
Get Started

Startup

€149

per month · 50M tokens included

All three models. Fixed monthly cost. Best for growing products with predictable AI spend.

  • All 3 models
  • 50M tokens/month
  • Overage at $0.10/1M
  • Zero prompt logging
  • EU data residency
  • Email support
Get Started →

Business

€399

per month · unlimited

Dedicated throughput. Predictable costs. For production workloads that can't share capacity.

  • All 3 models, unlimited
  • Dedicated capacity
  • SLA agreement
  • Zero prompt logging
  • EU data residency
  • Priority support
  • Data Processing Agreement (DPA)
Contact Us

Privacy & Compliance

Built for Europe's strictest requirements

Your data never leaves Germany. We built the infrastructure so GDPR compliance is the default — not something you configure later.

🙈

Zero prompt logging

Requests are processed in RAM and discarded immediately. We never store your prompts or completions.

🏠

EU data residency

All inference runs on bare-metal in Hetzner's German datacenter. No cross-border transfers, ever.

🚫

No training on your data

We run open-weight models. Your data is never used to train or fine-tune any model, anywhere.

📋

DPA available

Data Processing Agreement for Business customers who need contractual GDPR compliance documentation.

🇩🇪
Frankfurt, Germany
Hetzner Datacenter · EU Jurisdiction
GDPR Compliant EU Data Only No US Transfers Bare Metal AMD EPYC 128 GB ECC RAM

Integration

One line to switch

If your code runs on OpenAI, it runs on Cheshir. Change the URL and API key — nothing else.

Python
JavaScript
cURL
from openai import OpenAI

client = OpenAI(
  api_key="ck-your-api-key",
  base_url="https://api.cheshir.dev/v1"  # ← only this changes
)

response = client.chat.completions.create(
  model="qwen2.5-72b",
  messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "ck-your-api-key",
  baseURL: "https://api.cheshir.dev/v1"  // ← only this changes
});

const res = await client.chat.completions.create({
  model: "qwen2.5-72b",
  messages: [{role: "user", content: "Hello!"}]
});
console.log(res.choices[0].message.content);
curl https://api.cheshir.dev/v1/chat/completions \
  -H "Authorization: Bearer ck-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{'
    "model": "qwen2.5-72b",
    "messages": [{"role": "user", "content": "Hello!"}]
  '}'