AI models · data verified October 28, 2026

AI with EU data: which models suit companies under GDPR

Only Europe's Mistral keeps data in the EU by default. Google and OpenAI offer an EU region for companies, and Alibaba does for Qwen3.8-Flash. The direct Claude, Grok, DeepSeek and Kimi APIs do not guarantee EU processing, and Meta doesn’t mention any. The safest option is an open model on your own server.

  1. Mistral Medium 3.5 Mistral AI · balanced

    Mistral is a French company and stores data in the EU by default. Data only goes to the US when a developer explicitly uses the US endpoint, and some features may pass it temporarily to subprocessors. Medium 3.5 is the default model in the Vibe app, and the Team plan costs €24.99 per user before VAT. Watch out: on Free and Pro, conversations may be used for training until you switch that off.

    API price: $1.5 / $7.5 per 1M tokens (input / output) · Context window: 256K tokens

  2. Gemma 4 31B Google · small

    For GDPR, a model on your own server is the safest option: the data never leaves the company. Gemma 4 31B under Apache 2.0 runs on a single graphics card (about 17.5 GB at 4 bits) and handles more than 35 languages. Qwen3.8-27B and Muse Glimmer 30B work the same way. You do need someone to set up and maintain the model.

    API price: no published API price per 1M tokens (input / output) · Context window: 256K tokens

  3. Mistral Large 4 Mistral AI · flagship

    The largest model from a European company, trained in Mistral’s European data centres according to Mistral. In the API it has a 1-million-token context, data is processed in the EU by default, and it costs $1.36 / $4.18 per million tokens. It is still a preview, though, Mistral published most of the test results itself, and in Surge AI's blind coding evaluation it scored 3.74 out of 5 against 4.22 for Claude Opus 5.

    API price: $1.36 / $4.18 per 1M tokens (input / output) · Context window: 1M tokens

  4. Gemini 4 Argon Google · flagship

    On Gemini Enterprise Agent Platform (formerly Vertex AI) it runs in the eu region, so data can stay in the EU. Under its terms, Google does not train on API data from EEA users, and Google Workspace does not use company data for training. It ranks first of 43 models on the Vals Index. The introductory price of $2 / $10 doubles from December 2026.

    API price: $2 / $10 per 1M tokens (input / output) · Context window: 1M tokens

  5. GPT-6.1 Sol OpenAI · balanced

    OpenAI offers EU processing in the API for 10% extra, and the Enterprise plan has data residency in 10 regions including the EU. GPT-6.1 Sol scores 55 on the Artificial Analysis index and costs $2 / $10. ChatGPT Business does not use company data for training by default. You do have to choose the region explicitly and pay for it.

    API price: $2 / $10 per 1M tokens (input / output) · Context window: 1.1M tokens

  6. Qwen3.8-Flash Alibaba (Qwen) · fast and cheap

    The only main Qwen model that runs in EU scope in Frankfurt: data is stored in Germany and processing stays inside the EU, at the same price as in Singapore ($0.15 / $0.47). Qwen3.8-Max and Qwen3.7-Plus run there only in Global scope, with no EU boundary. The vendor is China's Alibaba, and the test results so far are its own.

    API price: $0.15 / $0.47 per 1M tokens (input / output) · Context window: 1M tokens

Who has no EU data residency

  • Anthropic: the direct Claude API processes data globally or in the US and stores it in the US, with no EU region to choose. Check the options with cloud partners such as Amazon Bedrock or Google Cloud.
  • xAI: it only guarantees the processing location in the US, through its US endpoint. Grok 4.7 and Grok Build 0.1 run in the US only. The older Grok 4.3 also lists the European eu-west-1 region, but xAI says its global API may move requests between regions. xAI did open Grok 4.5 to EU users in July 2026, but it also runs only in the US. Grok has no EU data zone on Microsoft Foundry either.
  • DeepSeek: the app, website and API store personal data in China. Czechia's NÚKIB rates the threat as high, and they are banned in Czech state administration.
  • Moonshot (Kimi): API data may be stored in Singapore, and EU processing is not offered.
  • Meta: for Muse Spark and Muse Wonder on the Meta Model API, Meta doesn’t mention any EU processing option or publish an official country list.
  • Alibaba: Qwen3.8-Max and Qwen3.7-Plus run in Frankfurt only in Global scope, so processing is not tied to the EU.

What else to sort out for GDPR

The EU region is only part of it. Also check whether your data is used for training: ChatGPT Free and Mistral Free and Pro do so until you switch it off. Business plans such as ChatGPT Business, Google Workspace and Mistral Enterprise do not use your data for training by default. Sign a data processing agreement with the vendor and do not send the AI more than the task really needs.

Tip: Mistral offers the Chinese models GLM-5.3 and GLM-5.3-Flash in its API with data in the EU, at the same price as Z.ai.

How we chose

The order reflects how reliably your data stays in the EU. First comes the European vendor that keeps data in the EU by default and offers a ready-made app and API. Right behind it is your own server, the safest option for data, but one that needs hardware and an admin. Then come US and Chinese services, where you have to choose or pay for an EU region. For each model we also weigh quality, price and training on your data. This is not legal advice: GDPR compliance also depends on your contract with the vendor and on what you send to the AI.

FAQ

Does ChatGPT store data in the EU?

If the company chooses it. ChatGPT Enterprise offers data residency in 10 regions including the EU, and in the OpenAI API you can switch on EU processing for 10% extra. For the regular plans, OpenAI does not list an EU region option.

Does Claude store data in the EU?

Not through the direct API. Anthropic processes data globally or in the US and stores it in the US. Claude works normally in Slovakia and Czechia, but companies that need data kept in the EU have to check the options with cloud partners such as Amazon Bedrock or Google Cloud.

Why do EU companies choose Mistral?

Mistral is a French company, stores data in the EU by default, and many of its models have open weights, so a company can also run them on its own server. On Enterprise, data is not used for training by default. Some features may temporarily pass data to subprocessors, which Mistral lists in its Trust Center.

Sources

Prices and models change often. The data comes from official vendor pricing pages and documentation, and every page shows when it was last checked. Before a larger purchase, confirm the price with the vendor.

← All AI models

Not sure which model fits your business?

I can help you pick a model and plan based on what you actually need, set it up safely for company data and train your team to use it well. Let’s start with an intro consultation.

See pricing jozef@walterstein.eu