Switch between gpt-oss-120b, Qwen and Ministral in the chat. Two models answer the same question side by side and you pick the better answer.
gpt-oss-120b
Qwen3.8-27B
Qwen3.6-35B
Ministral 3 14B
02Analyse documents
Drop in a contract, get deadlines and amounts
Upload PDF, Word, Excel or PowerPoint. The AI summarises, lists deadlines and amounts and translates on request.
PDF
Word
Excel
PowerPoint
03Create a knowledge base
Your own documents as a knowledge base, called up with # in the chat
Create a knowledge base, upload files, sync whole folders or add web pages and text. In the chat, a # brings in the collection and the answer cites its source.
Collections
Folder sync
Access per group
04Web search
Current information from the web, with sources to check
For publicly available information you switch on web search per chat and attach individual web pages directly. The answer cites its sources; only the search terms go to search engines, without reference to person or organisation.
Web search
Attach web page
Can be disabled per group
05Scans, images, speech
Read scans, capture dictation, listen to answers
Text recognition for scans and photos, image understanding, dictation by microphone and speech output. All with models operated in Germany by German providers.
Text recognition
Vision
Dictation
Read aloud
06Assistants
Recurring tasks as an assistant for each department
Set the instructions, model and knowledge base once. The team picks the assistant like a model, templates live in the prompt library.
Assistants
Prompt library
Templates
07Administration
SSO sign-in, permissions per group, usage at a glance
Sign in with Microsoft Entra ID, Google or Keycloak. Groups control models, knowledge bases and features, analytics show usage.
SSO
Groups
Usage analytics
Features
Included out of the box: more than a chat window
Every feature is set up and runs on the models operated in Germany by German providers.
Work
Chat with several models
Pick a model per task, compare two side by side.
Web search with sources
Current information from the web, every answer with sources.
Attach a web page
Paste a URL and summarise or query its content.
Upload files
PDF, Word, Excel and PowerPoint, even entire contracts in one go.
Analyses with charts
Python runs in the browser, for tables and charts.
Temporary chat
Chats without history for particularly sensitive questions.
Knowledge
Knowledge bases with sources
Answers from your own documents, with a source to check.
Maintain collections
Add files, whole folders, web pages and text.
Assistants per department
Instructions, model and knowledge as a ready assistant.
Prompt library
Call up proven templates with a slash.
Folders for projects
Organise chats per project, client matter or case.
Notes and earlier chats
Refine notes with AI, use chats and notes as context.
Image and speech
Text recognition
Scans, photos and forms become readable.
Image understanding
Describe and analyse photos, charts and screenshots.
Dictation and transcription
Speak instead of typing, meetings as text.
Voice mode
Talk to the AI hands-free, answers as speech.
Read aloud
Answers as speech output.
Translate
Texts and documents into the common European languages.
Team and administration
SSO sign-in
Microsoft Entra ID, Google or Keycloak, with multi-factor.
Groups and permissions
Who uses which models, knowledge bases and features, web search included.
Channels and sharing
Share chats, work with the AI together in a channel.
Personal instructions
Set role, tone and memories per user.
History and devices
Search chats, save them as PDF, in the browser on any device.
Usage analytics
Messages, tokens and models at a glance.
Models
These models are included and selectable in the chat
Open-weight models operated in Germany by German providers, without US clouds and without training on your inputs. As of October 2026.
Model
Use
Context
gpt-oss-120b
Text, analysis and tool calling
131,000 tokens
Qwen3.8-27B
Text, images and multi-step tasks
256,000 tokens
Qwen3.6-35B
Long documents and image understanding
256,000 tokens
Ministral 3 14B
Fast chat with vision
262,000 tokens
GLM-OCR
Text recognition in scans, PDF, Word, Excel and PowerPoint
131,000 tokens
Qwen3-Embedding-8B
Search across your knowledge bases
32,000 tokens
Qwen3-VL-Reranker-2B
More precise hits in knowledge bases
32,000 tokens
Whisper large v3 turbo
Dictation and meeting transcription
–
Qwen3-TTS
Read answers aloud
–
§203 StGB
A contract chain in which every link is bound
For professional secrets, everyone involved must be bound to confidentiality. Our chain is complete.
Your team
Sign-in via SSO with your multi-factor authentication, permissions per group.
SSO
MFA
WZ-IT
Setup and operations, DPA and Section 203 supplementary agreement, staff bound in writing.
DPA
§203
Instance
Dedicated instance per organisation, no shared platform; data centres in Germany run by German providers, with DPA and Section 203 supplementary agreement.
Germany
DPA
§203
Models
AI hosting by a German provider in Germany, no transfer abroad, no US provider, no training.
Germany
DPA
§203
Operating models
Data storage on your terms: hosted, in your network or fully on premises
All three options use the same interface and no shared platform. You decide where the platform, chats and models run.
You pay a fixed price per organisation and month with a set number of users included. Additional users cost €14.90 each; if the next package is cheaper, we switch. Fully used, that equals €14.99 (Start) to €12.00 (Organisation) per user, models and contracts included.
Dedicated instance on your own subdomain or domain
All models, knowledge bases, assistants and text recognition
Hosting in German data centres of German providers, no US clouds
DPA and Section 203 supplementary agreement
Updates, daily backup and 24/7 monitoring
Next-business-day support, various service levels available
Optional onboarding package for €990 once: knowledge base with your documents, assistant templates and two hours of training documented under Article 4 of the AI Act. Additional users €14.90 each per month, up to 150 users in the Organisation package; beyond that or with dedicated GPUs a custom offer. Cancel monthly.
All prices are net and exclude statutory VAT. The offers are addressed to businesses.
Process
Your own AI platform in four steps
You test with your team first, then decide.
01
Initial call
Clarify users, tasks, sign-in and requirements.
02
Test instance
Test for 7 days free of charge with non-critical data; the test instance is not Section 203 compliant.
03
Quick start
Within a few working days: set up instance, domain, users and SSO, plus a one-hour walkthrough.
04
Operations
We handle updates, backups, monitoring and new models.
For your industry
The platform for your industry or department
Use cases, legal framework and questions per audience, plus the options with your own hardware.
Prices, Section 203, models, data protection and process
An AI platform combines chat with several language models, knowledge bases from your own documents, assistants and permission management in one interface for all staff. Unlike private chat accounts, the organisation controls sign-in, access and storage location. The WZ-IT AI platform runs as a dedicated instance based on Open WebUI, set up and operated by WZ-IT.
An AI tool solves one task, such as transcription or translation. An AI platform bundles several models and features for all staff: chat, knowledge bases, assistants, web search, dictation and permissions, with central sign-in and one contract for the whole organisation.
It covers the tasks teams usually use ChatGPT for: drafting texts, summarising documents and asking questions about your own files. On data protection, it relies on a DPA, processing exclusively in Germany by German providers and no use of inputs for training. Your data protection officer assesses your specific use; this is not legal advice.
On multi-tenant platforms many organisations share the same application and database, separated only logically. With WZ-IT there is no shared platform: every organisation gets its own instance on its own domain, with its own database, knowledge bases, groups and settings. Updates, backups and permissions affect only your instance.
Start with 10 users costs €149.90, Team with 25 users €399.90, Business with 50 users €699.90 and Organisation with 100 users €1,199.90, each net per month. Additional users cost €14.90 each per month; if the next package becomes cheaper, we switch. The quick start with SSO and the Section 203 supplementary agreement are included even in the smallest package. The optional onboarding package costs €990 once. You pay for the package per organisation; only users beyond the included number are charged individually. Fully used, that works out at €14.99 (Start) down to €12.00 (Organisation) per user. For more than 150 users we prepare a custom offer.
No. Section 203 StGB additionally requires every contributing service provider to be bound to confidentiality. We therefore conclude a Section 203 supplementary agreement in addition to the DPA, bind our staff in writing and hold the same supplementary agreement with our hosting providers, the data centre of the instance and the AI hosting of the models. Whether a use case is permitted under your professional rules is for you to assess; this is not legal advice.
Yes. The instance can be set up so that it is reachable only from your network or via VPN. On request it runs on premises, for example on a WZ-IT App Node at your company that we also operate. The models then still come from a German AI hosting provider in Germany; for models on your own premises there is the AI Cube.
In your own instance in a data centre in Germany run by a German provider with ISO 27001 and BSI C5: chats in each user's account, documents in the knowledge bases, plus a daily backup. The AI hosting, also a German provider, processes requests in Germany and does not use them for training. When the contract ends, the data is deleted or returned in accordance with the DPA.
No. The models run with a German AI hosting provider in Germany that does not use inputs for training. By default, chats are stored in each user's account within your instance, and knowledge bases also stay in your instance; for matters without history there is the temporary chat.
Exclusively in Germany with German providers: the platform instance with WZ-IT in a German data centre, the language models with a German AI hosting provider. For documents, chats and models there is no transfer to third countries and no US provider in the data path. When you use web search, only the search terms go to search engines, without reference to person or organisation.
Yes. For publicly available information you switch on web search in the chat or attach a web page directly; the answer cites its sources. Only the search terms go to search engines, without reference to person or organisation, and documents and chats stay in Germany. Administrators disable web search per group, for example for teams handling professional secrets.
gpt-oss-120b, Qwen3.8-27B, Qwen3.6-35B and Ministral 3 14B for chat and writing, plus embedding and reranker models for knowledge bases, GLM-OCR for text recognition, Whisper for dictation and transcription and Qwen3-TTS for speech output (as of October 2026). We enable new models and replace retired ones in good time.
These models run with US providers, which would make a US company part of the processing chain and calls for an additional assessment for professional or official secrecy. We therefore use open-weight models operated in Germany by German providers. They work well for drafts, summaries and questions about your own documents; the test instance lets you check this with your tasks.
The allowance is designed as fair use for the typical usage of the package. We inform you before it is reached and discuss the next package if usage stays higher.
Yes, through knowledge bases. You upload documents and the platform answers questions with a reference to the source. You decide which group sees which knowledge base. With the onboarding package we set up the first knowledge base together with you.
Single sign-on with Microsoft Entra ID, Google or Keycloak is included in all packages, each with your multi-factor authentication. You upload documents into knowledge bases. Connecting SharePoint, CRM, ERP or ticket systems is not part of the standard scope and is planned as a separate project.
Whether a data protection impact assessment under Art. 35 GDPR is required depends on the data and purpose; your organisation decides this with its data protection officer. For the assessment you receive the DPA with the technical and organisational measures, and the contract chain on this page shows the data path. This is not legal advice.
Open WebUI is the platform's open-source interface; we run it with its original interface. The AI platform adds models operated in Germany by German providers without US clouds, the token allowance, SSO, operations, updates and the Section 203 supplementary agreement across the whole chain.
Since 2 February 2025, Article 4 of the AI Act requires companies and public bodies to ensure sufficient AI literacy among their staff. The onboarding package includes two hours of training, which we document.
The platform can be cancelled monthly with 30 days' notice to the end of the month. Beforehand we set up a test instance that runs for 7 days free of charge and without payment details. The test instance is not Section 203 compliant and is meant for non-critical data; the Section 203 supplementary agreement applies from the smallest package, Start. Public bodies can choose a 12-month term with invoicing.
Within a few working days. After the initial call we set up the instance, domain, users and SSO sign-in; no hardware of your own and no in-house model operation are needed. Beforehand your team can use the platform free of charge for 7 days in a test instance.
For health data we recommend the AI Cube, an on-premises AI server. Under Section 393 SGB V, cloud services for health and social data require a C5 attestation.
Reviews & projects
What clients say about working with us
From local AI integration to architecture, data sovereignty and ongoing operations.
WZ-IT moved our studio infrastructure from decentralised individual devices to a central platform: every site is securely connected via VPN, new devices are onboarded automatically and an entire site is provisioned from a template, without manual steps on location. What impressed me most is the breadth and depth of their knowledge: Timo and Robin are not a typical IT provider who sets up a server and leaves. The two of them think their way into highly complex infrastructure and software topics, work through every requirement we put in front of them, and build networking, provisioning and operations so that everything fits together in the end. WZ-IT is an excellent partner for complex software, network and architecture projects.
Steve KirchnerManaging Director, nextGYM GmbHView project
International
Built in Germany's Ruhr Valley. Running worldwide.
WZ-IT designs, develops and operates infrastructure and software for clients in Germany and internationally. We deliver projects remotely and continue supporting them in ongoing operations after go-live.