One AI chat.
All models.
GDPR-compliant.

The world's frontier AI – plus our own models on European inference.

GPT, Claude, Gemini, Mistral, Perplexity, Kimi – plus anymize Waterfall and Fountain. Automatically anonymized when you need it.

The switch

You have eight tabs open.
We have one interface.

Every day you pick: ChatGPT for the first draft. Claude for the precise analysis. Gemini for research. Perplexity for sources. Five tabs. Five accounts. Five subscriptions. Five headaches when you're handling sensitive data.

anymize ends the tab chaos. One chat. All models. One interface. Automatically anonymized when you have sensitive data. And when you don't, it takes the direct path – no detour, no policy theater.

What that changes: You switch models mid-conversation when the question changes depth. You pull context from files, YouTube transcripts and your own knowledge bases without copy-paste. You dictate instead of typing when it needs to go faster. And you have a single place where you see compliance – not five.

All models

The world of AI.
In European hands.

There are many "GDPR-compliant AI chats." Most host an EU server and call that "data protection compliant" – but still send their customers' original data to the same US models. A European server with an American model that sees your client data unfiltered is legally no different from using ChatGPT directly. The server is not the problem – it's the unencrypted data that leaves it.

anymize solves this differently. Twin-track.

Track 1International frontier models

International frontier models – GDPR-compliant through anonymization

The world's most powerful AI models – in their current version, at full performance. Not stripped-down "EU versions," not older generations:

OpenAI (GPT)
Broad general knowledge, text processing, code
Anthropic (Claude)
Legal analysis, long documents, precise argumentation
Google (Gemini)
Multimodal tasks, up-to-date research, image understanding
Mistral
European provider, strong all-rounder
Perplexity
Web-focused research with source citations
Moonshot AI (Kimi)
Deep analysis with very large context windows

Because your data is automatically anonymized before it goes to any of these models, none of these providers ever see your sensitive original content. You use them at full strength – and remain fully GDPR-compliant. Not because the server sits in Europe. But because nothing sensitive leaves the EU.

Track 2anymize's own models

anymize's own models – European inference, no need for anonymization

Not every task tolerates anonymization. Code loses structure when variable names are replaced. Excel formulas break when cell references become placeholders. JSON becomes inconsistent when keys are anonymized. For these cases, we operate two in-house AI models, fully in the EU, under the control of GRVITY GmbH (Kiel, Germany).

Both models are based on the latest Qwen architecture and additionally fine-tuned by us on a German-language professional dataset (legal, medical and business content). Two variants, one stack:

Deep thinking

anymize Waterfall

Our deep-thinking model

Our most powerful in-house model for deep analysis. Ideal for complex analyses, multi-step planning, structured problem-solving – when you need frontier-level quality but the data cannot leave the EU.

Performance

anymize Fountain

Our performance model

Fast throughput, very strong coding capabilities. Ideal for writing and refactoring code, Excel/CSV analyses, JSON and API processing, SQL work, structured data extraction – everywhere where fast and precise answers on structured data matter.

Performance positioning – honest

The most powerful open-source models are typically one version behind the proprietary frontier models. Close enough to handle demanding tasks – but honestly, not state of the art. Our advantage: because we fine-tune every new open-source release on a German-language professional dataset, our models are often more competitive in German-language professional contexts than pure benchmark numbers suggest (useful across German-speaking Europe: Germany, Austria, Switzerland). And as soon as a new, stronger open-source generation appears, we follow quickly – one run through our pipeline, and the new level is available in anymize.

Why both tracks together are strong

Because inference, data storage and operation of Waterfall and Fountain happen entirely with us, the contracts signed with you apply directly: the DPA under Art. 28 GDPR and – for professional secrecy holders – the consent statement under § 43e BRAO (a German lawyer-confidentiality statute; equivalent frameworks apply across most European jurisdictions). No additional anonymization needed. No third-country transfer. No further DPA. The response quality is sufficient for most structured and domain-specific tasks – and for everything else, the international frontier models are available with anonymization.

In preparation

anymize Spring

A third in-house model is in development – anymize Spring – optimized for ultra-fast queries and high throughput. For scenarios like real-time classification, simple text transformations or fast previews. Available soon.

Performance classes – selectable per question

Both tracks offer three performance levels:

Deep analysis

Deep

For complex analyses, strategic decisions, deep reasoning tasks

Everyday

Smart

For everyday use: balance of quality and speed

Fast

Fast

For fast answers, simple tasks and high volumes

Switch with one click – mid-conversation. Between tracks, between providers, between classes. Without leaving the chat. Without losing context.

New models arrive with us quickly. As soon as a provider releases a new frontier model, it is usually available within a few days – without you having to create a new account, sign a new subscription or learn a new UI. The same applies to our own models: Waterfall and Fountain receive regular updates, while the billing, compliance and integration layer remains stable for you.

Upload content

Everything the
answer needs.

An AI chat is only as good as the context it knows. anymize lets you bring five types of content into the conversation – directly from the chat, without detours:

01

Upload file (anonymized)

Your sensitive documents. PDFs, Word, Excel, images, scans. Automatically anonymized before sending – 40+ data categories, over 95% detection rate. The AI responds with full context; you get the result back with your original data.

02

Upload file (not anonymized)

For non-critical content. Publicly available studies, internal manuals without personal references, product descriptions. Direct path – no anonymization detour when not needed.

03

Fetch YouTube video

Paste the URL once, and anymize pulls the video's transcript into the chat. Ideal for expert talks, competitor analyses, investor calls. The model summarizes, extracts arguments, compares with your other sources.

04

Fetch website content

One URL is enough. anymize reads the page, renders JavaScript-driven content and pulls it into context. Correctly cited. Ideal for pitch research, due-diligence source texts, market observation.

05

Existing document

Direct access to documents from earlier chats or projects, without re-uploading them. Especially useful when you're working through a series of sessions on the same case.

Plus: voice input. The microphone button starts a self-hosted speech recognition – understands domain language, with strong support for European languages (including German). Dictating is often faster than typing, especially in meetings and notes.

Tools & reasoning modes

Thinking modes
that match the question.

01

Reasoning modes

How deeply should the model work? You decide per message:

Fast

Direct answer without internal loops. For quick questions and high volumes.

Thinking mode

Structured reasoning before the answer. For problems that need more than one step.

Max

Detailed reasoning, longer context windows, maximum care. For analyses where every nuance matters.

02

Additional capabilities

Intensive research

Multi-step research with source analysis. anymize searches structured information, verifies sources, compares statements and delivers a reliable result. Ideal for due diligence, market briefings, preparation notes.

Web search

Up-to-date information from the web, right in the conversation. Live sources, correctly cited.

Generate image

Visualizations, infographics, concept sketches. From your description, directly in chat.

Connectors

Beta

Tools like HubSpot right in chat. Retrieve customer data, create tickets, prepare emails – without switching interfaces.

Knowledge bases & placeholders

Your knowledge.
In context.

Knowledge bases in chat

Contracts, policies, project files, minutes – bundle them in a knowledge base and activate them in chat with a click. The AI answers with reference to your internal sources, not just its world knowledge. No manual uploading for every conversation, no context loss.

Perfect for recurring topics: one knowledge base per client, per project, per domain. Access granularly controllable – some databases are private, others team-wide.

@

@ for placeholders

Type @ in chat to insert a named placeholder – for example @Client, @ContractDate, @Amount. Unbeatable when you're building templates, testing prompt chains, or reusing contexts across cases. Placeholders are filled before sending – manually or automatically from a dataset.

Anonymization in chat

Secure when you need it.
Direct when you don't.

For every upload and every prompt, your workspace (or you) decide whether anonymization applies:

For sensitive documents, names, addresses, IBAN, case numbers, health data and 40+ more categories are automatically detected and replaced with placeholders before the content reaches any AI model.

For non-critical content, the direct path – no detour, no policy theater.

Bidirectional anonymization ensures that the response ends up containing your real data again. No AI provider ever sees the original.

For teams and companies, document or prompt anonymization can be enforced by admins – including four-eyes review and a 12-second review countdown. So compliance stays policy, not hope.

What you do with it

When you want quality
fast.

Seven scenarios from real professional practice. No demos, no show cases – what anymize customers actually do:

01

80-page contract understood in 30 minutes

Upload (anonymized) · Model on Deep · Ask about "clauses with liability risk" · Activate cross-references in your knowledge base · done.
02

Due diligence in one hour

Upload three years of annual reports · Intensive research activated · Ask for "anomalies, trends, inconsistencies" · Export short report as artifact.
03

Discharge letter in 20 minutes

Upload findings scan · Pull clinic guidelines from your knowledge base · Thinking mode · Draft to specification, directly editable.
04

Competitor analysis from YouTube + web

Paste interview URL · Fetch company website · Complement with Perplexity or Web search · Generate executive summary.
05

AGG-compliant applicant screening

200 CVs (anonymized) · Criteria as placeholders · Ranking via Max reasoning · Shortlist in minutes.
06

Customer email with CRM context

Activate HubSpot connector · Customer history in chat · Specify tone · Draft email ready to send.
07

Code refactoring with company standards

Paste code from your IDE · Switch to anymize Fountain · Internal coding guidelines from your knowledge base · "Refactor to our conventions" · Output goes straight back. No anonymization needed – European inference, no data leaving the EU.

What you should know about anymize Chat.

Frequently asked questions

anymize Chat is the multi-model chat interface for professional users. Instead of taking out individual subscriptions with OpenAI, Anthropic, Google, Mistral and Perplexity, you use all the frontier models in one interface – with automatic anonymization when you have sensitive data. Developed in Germany, hosted in the EU.

Start now.
14 days free trial.

All models. All features. No credit card.

We stand behind anymize. And we know – when an AI tool touches client, patient or employee data, a demo video isn't enough. That's why we give you 14 days of full access – all models, all features, no credit card. Enough time to be certain, before you trust us.

Your AI workplace awaits.