Skip to content

Blog · Published · Updated

Private LLMs: data, EU hosting and on-premise

Short answer

A private LLM lets you choose where the answer is computed, but the app has more parts: text extraction, the index, logs and backups. Before you choose EU hosting or an on-premise server, draw a data map from the document to the answer. At Syntalith an app on a local model costs from €3,000 net plus GPU hardware. At 400 questions to documents a month of 12 minutes each, it pays back in 1.6 months.

How we workDetails ↓

  1. 01Call30 minutes with an engineer, free
  2. 02PrototypeA working system on your process and data, free
  3. 03ImplementationTarget and fixed price in the contract; 50% at signing
  4. 04Acceptance50% once the agreed target is met
model on your server or in the cloud in the EU
quoted after the review
same work in application with EU cloud model
from €3,000
task, data and equipment analysis
€0

We belong to the OpenAI and Anthropic partner programmes and learn through their official courses. What this means ↓

Author: Syntalith · 3 min read

Example: the morning request inbox
Agent · ready

Statistics

Private LLM and data: statistics 2026

1.6 mo.

payback of a €3,000 local app at 400 questions to documents a month: €2,792 of work today, €912 with checking and the GPU server. Scenario.

Figures checked: 22 October 2026

A server in the office, logs in the cloud

„The data stay with us” can mean the company's building, its cloud account, an EU region or a ban on sharing content with the model provider. Each answer leads to a different architecture and cost. On-premise describes a location and an air gap describes isolation from other networks, and the two are not interchangeable: a server in the office can send logs to a cloud service if monitoring is set up that way.

A document passes through text extraction, search, the model and the app, and each stage may leave temporary files and logs. The model that creates the document's representation for search can also run locally or as an external service. An agent's tools, such as a call to the CRM or an internet search, are separate data flows.

A data map for a private LLM
  1. 01Documenttext extraction locally or in a service
  2. 02Indexpassages with the user's permissions
  3. 03Modelwhere the answer is computed
  4. 04Answerthe chat record and how long it is kept
  5. 05Backups and logswhere they are and who reads them

What it costs and when it pays back

Scenario: 400 questions to documents a month
Today
400 questions to documents × 12 min = 80.0 h (€2,792)
With the app, an employee checks the output in 25% of the time
€698
Hetzner GEX45 GPU server
€214
Saving
€1,880 a month

Payback of €3,000

1.6 months

The volume, the time and the checking share are the scenario's assumptions, which we measure in the trial. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat. Price of the GEX45 server with 24 GB of GPU memory from Hetzner's page on 22 October 2026, without the one-off €209 fee.

The count covers the time spent searching, reading and compiling material. What remains is the time to check the result against the source, 25% of today's work in the scenario. We measure both times in a trial on 20 to 100 cases the team has already handled.

Permissions must work before the answer

A salesperson and a lawyer may use the same model but see different documents, so the app limits search by the user's permissions. An instruction in the prompt does not replace access control. We also test a change of role, the removal of permissions and the deletion of a document, because the index, the cache and the chat record may each need a separate update. A model fine-tuned on confidential examples needs protection too: we limit access to the training set, the adapters and the weights.

What to agree with IT beyond the model itself
  • Where the app takes documents from and where it saves results
  • What stays in logs, backups and the history of questions
  • Who has access to questions, answers and outputs
  • Who updates the model, the app and the document set

When EU hosting is enough

If the requirement is an EU region and a data processing agreement, a service in a European cloud region often costs less than your own server. When the requirement forbids internet connections during operation, you need a server on site or an isolated network. The GDPR requires a legal basis for processing, data minimisation, limited retention and security measures, and processing outside the EEA is subject to the transfer rules. The person responsible for data protection assesses these questions, and Syntalith provides the technical description, the data flow and the configuration. Signing off an environment with no connection is described in a local model in manufacturing.

An EU region, a GPU server or your own server room
  • An EU region and a processing agreement are required

    a service in an EU region

    no hardware

  • A model without sending content to the provider

    a GPU server, for example the GEX45

    €214 a month

  • No internet connections

    a server on site and an isolated network

    priced after an analysis

Where to start

The most useful inputs are specific requirements: the allowed region, no connections during operation, user roles, the retention period and recovery after a failure. The phrase „the highest security” sets none of these parameters. We deploy local models as part of private LLMs, and we build the view for working with sources as an AI app. For the free analysis, a description of one case and of where the materials sit is enough. Confidential files stay in the company, and you get a written plan within 2 working days.

Worked example · Scenario

How we count the payback: a local model

We run the same count for your company in the free analysis.

€5,817

a month for the work on 1,000 questions

EU cloud version payback: 0.6 mo.

Our assumptions are 10 minutes to find the right instruction, and checking in 15% of the old time. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat, for companies with 10 or more employees. Model cost: Claude Sonnet 5.5, about 6,000 + 500 tokens per answer in the EU cloud version, at the ECB rate of 1.1205 USD per euro on 22 October 2026. We count the payback for an app with a model in the EU cloud (from €3,000 net). We price the model on your server after a call, and the cost of people's work stays the same.

See the full calculation →

This is how we implement it in companies

All case studies →

Client results.

All projects →
foap.comcustomer service

foap.com customer service: 95% of replies are sent automatically.

Users from several countries write to foap.com, in several languages. The system reads the entire conversation and the terms of that person, and then sends a response in that person's language. Questions about money and exceptions go to the team with a ready draft, so they have to check and send them.

the system sends the answers itself, without human typing
95%
time from the arrival of the e-mail to the ready response (median)
30 s
e-mails that require a response receive a ready-made draft
100%
less work for every 1000 answers (estimate)
47 h

Data from foap.com production, October 20-21, 2026. Estimate: 3 minutes for handwritten response.

How foap.com customer service works →
Sample messages. Money matters always come to the person.

Other implementations

jaktako.design

studio projektowe

−87%

Time to look for new customers, every day

4 hours → 30 min

approx. 73 hours less per month (21 days of work)

according to the studio

How the agent loop works →
Signature Estates

real estate

approx. −90%

Operator time spent browsing portals, per day

2–3 hours → 20 min

approx. 45 hours less per month (21 days of work)

respect · contact the seller 5 minutes after the announcement

How the agent finds property listings →
Kuba Koziej, CEO and co-founder of MoreGrowth and board member at Natu.Care

Kuba Koziej

CEO and co-founder, MoreGrowth

„We started with my own AI agent. During the training that followed, I learned to build new applications myself. Today I am bringing this way of working with AI into my business. Working with Syntalith gave me exactly the skills and tools I needed.”

View the original training review on LinkedIn ↗

A model on your server or in the cloud in the EU, proven on your tasks.Book a free process review (30 min) →

How we build automation and agents

From a conversation to a working system in four steps.

  1. 01 · 30 min

    Free process review

    €0

    You're talking to an engineer. You receive the written plan within 2 business days.

  2. 02 · 5 days

    Free prototype

    €0

    The system works on 20-100 cases that your team has already handled. You see every result.

  3. 03 · 2–16 weeks

    Build

    50%

    Fixed price and goal written in the contract. First installment upon signing.

  4. 04 · after reaching the goal

    Reception

    50%

    You pay the second installment when the system reaches its target on your cases.

How a free prototype is created

Mon.

We set a goal

What result do you consider good?

Tue–Wed

The system is working

It goes through your affairs, one by one.

Thu

We compare

We check every difference with the team's work.

Fri

Result and price

List of errors and fixed implementation price.

You choose the matters. We anonymize them upon request, they stay in the EU and we delete them after the test.

On the market, such a test usually costs from PLN 3,000 to PLN 8,000.

Book a free process review (30 min)

We establish a prototype during a free analysis.

Data checked on October 25, 2026.

„Is my data safe?”

Yes. You choose where they go: from your company account to the model on your server.

  1. 00Employee's private account

    Customer and company data ends up in an account that the company cannot see or control. There is no data processing agreement, so it is a violation of the GDPR. When an employee leaves, the chat history goes with them.

  2. 01ChatGPT or Claude corporate account

    The provider does not use your data to train models. A copy of the queries remains in his logs for up to 30 days.

  3. 02API with zero retention (ZDR)

    The provider processes the request and does not save anything. It requires his consent, we prepare the application together.

  4. 03Processing in the EU

    Queries are counted and stored in Europe: OpenAI in the EU region, Claude via clouds with the EU region.

  5. 04Model at your place

    The open model runs on your server or in your cloud in the EU. The data does not go outside the company.

What we do with every implementation

  • Confidentiality agreement and data processing agreement before we receive any data.
  • You choose the cases for the prototype and we anonymize them upon request.
  • The data stays in the EU and we delete it after the test.
  • Corporate accounts with an administrator on your side.

When data cannot leave the company

The open model runs on your server or in your cloud in the EU. To choose from:

White-tailed eagle
Polish model, SpeakLeash and ACK Cyfronet AGH
PLLUM
Polish model, HIVE consortium commissioned by the Ministry of Digitization
Qwen
Alibaba
GLM
Zhipu AI
DeepSeek
DeepSeek

We select a model for the task and check it on your cases. We run it, tune it and maintain it.

Source: Bielik and PLLuM, ACK Cyfronet AGH ↗ · PLLuM, Ministry of Digitization ↗

Do you have a lot of your own data? We will fine-tune the small model.

  1. 1 Your examples
  2. 2 tuning
  3. 3 test on new cases
  4. 4 model at your place

When you have thousands of similar emails, documents or reports, we teach a small open model based on your examples. It works for you, responds quickly and costs less to maintain than a large model. We measure quality on issues that the model has not seen while learning.

Przemysław and Piotr did it in large programming and consulting companies. Artem did it before and does it today at Syntalith. It's about models on your own infrastructure, tuning and quality measurement.

You choose the protection level during the interview. We sign the data processing agreement before we see any data.Schedule a free analysis →

Private AI models

A model on your server or in the cloud in the EU, proven on your tasks.

We select the model and equipment for the task, run it and measure the quality using your examples before you buy anything. The data stays in the company or in the cloud in the EU. We price implementation, hardware or hosting and support separately after analysis.

View private AI models →
model on your server or in the cloud in the EU
quoted after the review
same work in application with EU cloud model
from €3,000
task, data and equipment analysis
€0

Questions

  • How does on-premise differ from EU hosting?

    On-premise is a server in the company's building, EU hosting a server in a European data centre. Both can send logs or backups elsewhere, so the data flow map and the configuration decide how the data are protected.

  • Is a local model enough for GDPR compliance?

    No. Using local weights or a server in the EU does not settle whether the process complies. You have to set the legal basis, the scope of data, the retention period, the permissions and the contracts.

  • How much does an app on a local model cost?

    At Syntalith from €3,000 net plus GPU hardware, for example a Hetzner GEX45 server with 24 GB at €214 a month. At 400 questions to documents a month the payback is 1.6 months.

Let's talk

Just write, call or choose a date.

We work fast because we hate corporate cotton wool. We don't have salespeople or managers to keep an eye on managers. You talk immediately to engineers who then build your system. We respect each other's time and yours too.

We will accept 3 more companies by the end of the year.

Mon–Fri 9:00–17:00 CET

What our OpenAI and Anthropic partnerships mean for you.

OpenAI makes ChatGPT and Anthropic makes Claude. We belong to the partnership programs of both companies, so we work with their tools every day and learn from the creators themselves.

OpenAI Select PartnerClaude Partner Network
01We will select a tool for the task
We implement both ChatGPT and Claude. We check both on your documents, and you choose the one that makes fewer corrections and saves more time.
02Knowledge straight from the creators
We learn from official OpenAI and Anthropic courses and materials. Your team learns about the features and settings that work today.
03Licenses without our margin
When we roll out ChatGPT, Claude or Copilot, you buy the licences directly from OpenAI, Anthropic or Microsoft and they invoice you. We add nothing to the licence.
04Certified engineers
On the team, we have an AWS certification in machine learning and an Anthropic certification in systems architecture on Claude.

Participation in programs does not mean that OpenAI or Anthropic endorses our services.