Model provider · in regular use
Mistral AI agency in Nice
A French vendor whose models can run in Europe, or even on your own infrastructure.
What we do with it
Mistral AI at our clients, concretely
On Ordelya, the restaurant management platform, the Mistral API translates menus, alongside DeepL and LibreTranslate. On LLM Monitor, its models and agents are among the AIs queried to measure brand visibility.
The limits
When we recommend something else
On a task where a larger model does markedly better, the sovereignty argument alone is not enough.
Our work
Projects we delivered with Mistral AI
LLM Monitor
SaaS that measures how visible brands are in AI answers.
Mistral AI: Mistral models and agents among the AIs measured.
Read the LLM Monitor case studyOrdelya
Restaurant management platform: back office, tablets, kiosks, screens and website.
Mistral AI: Automatic menu translation.
Method
How we bring Mistral AI into a project
In Ordelya, the restaurant management platform we develop, the Mistral API translates the menus shown on the website, tablets, kiosks and screens. The mistral-large model sits alongside DeepL and LibreTranslate, two other translation engines: the application does not rely on a single service, and each engine can be chosen according to the content being translated. In LLM Monitor, the service that measures how brands appear in AI answers, Mistral's chat models and agents are queried through LangChain, alongside OpenAI, Claude, Gemini and Perplexity.
For a client, we start from the use case and how sensitive the data is. Mistral offers an API with an option for hosting within the European Union, and some of its models are released under an open licence, which means they can run on your own infrastructure. During scoping we check which plan and contract actually apply, rather than assuming where processing takes place.
As with any provider, we compare models on a sample of your content, pin the chosen version in the configuration and open the account in your name, with separate keys for test and production.
For your business
What it changes in practice
For a small business subject to the GDPR, Mistral's appeal lies mainly in where processing happens. A French vendor, an offer that can be hosted in Europe and open models you can run yourself widen the options when data must not leave your perimeter. This is not an automatic guarantee: compliance depends on the plan in place, on what is sent to the model and on human approval downstream, all of which we address from scoping onwards.
The sovereignty argument is not enough on its own, though. On a task where another model gives markedly better results, we tell you so, and the decision is made with that in mind. In Ordelya, placing Mistral alongside other translation engines means choosing according to the content without tying the application to one provider. And as with any AI project, we start nothing without a use case costed in hours saved per week.
Decision guide
Should you choose Mistral AI?
The Mistral API or an open model running on our own infrastructure?
The API spares you buying and maintaining servers, and an offer hosted within the European Union can be enough for many tasks. Running an open model yourself ensures no data leaves your perimeter, but it needs suitable hardware, ongoing updates and often means accepting a smaller model. We decide based on how sensitive the data is and the quality achieved on your own documents, tested before any commitment.
Mistral or an American model for our project?
We compare both on your real use case, not on a leaderboard. A sample of your documents is put to the models under the same conditions, then we look at accuracy, cost and where processing takes place. If the data is sensitive and Mistral gives satisfactory results, it wins. If another model does markedly better and the data allows it, we recommend that one instead.
Mistral or a dedicated translation service?
In Ordelya we did not have to choose: the Mistral API, DeepL and LibreTranslate sit side by side to translate menus. A dedicated service such as DeepL is built for translation; a language model like mistral-large can also follow instructions on tone or trade-specific vocabulary. LibreTranslate can be installed on your own servers. Keeping several engines behind one function lets you choose according to the content and avoids lock-in.
Client questions
What clients ask us about Mistral AI
Does Mistral guarantee our data stays in Europe?
Not as a matter of course: it depends on the plan and contract in place. Mistral offers hosting within the European Union, and its open models can run on your own infrastructure, which settles where processing takes place. We review the terms in force with you during the project, and record the chosen execution mode in the processing documentation.
Where do you use Mistral today?
In two projects. Ordelya, our restaurant management platform, uses the Mistral API and the mistral-large model to translate menus, alongside DeepL and LibreTranslate. LLM Monitor, which measures how brands appear in AI answers, queries Mistral's chat models and agents alongside other providers such as OpenAI, Claude, Gemini and Perplexity, so the answers can be compared on the same question.
Is using Mistral enough to comply with the GDPR?
No. Choosing the provider settles one part of the question: where processing happens. You still need to limit what is sent to the model, plan for human approval when the output has consequences, and document the processing: records of processing, informing the people concerned, and an impact assessment where required. We address these from scoping and hand you the documents in writing.
Are Mistral's models as good as the others?
It depends on the task, and we would rather measure it than claim it. We put a sample of your documents to several models under the same conditions and compare the output with the expected results. On some tasks Mistral performs satisfactorily; on others a different model does markedly better, and we tell you so.
Where do we start?
With a free first conversation, in which you describe the task you have in mind and the kind of data involved. Together we check whether the gain can be costed in hours per week and which execution mode your data requires. You then receive a written, costed action plan. Whichever model is chosen, the code developed belongs to you.
Going further
Mistral AI sits within our Artificial intelligence
A technology does not make a project. The method around it decides the outcome: scoping, prototypes, short cycles, and a clean way out.
Neighbouring technologies
A Mistral AI project in mind?
The first conversation is free. We will tell you plainly whether this technology is the right one, and if not, which is.