That is, even more retarded than Proton’s Lumo, which is opaque as to which LLMs are used, and unbelievably dumb. But the Swiss persevere: Euria: the free, sovereign AI assistant to no longer depend on the American giants. They even pretend to heat homes: “The heat produced by Euria is fully recovered and reinjected into Geneva’s district heating network.”

But they’re retarded on two counts.

A retarded IT infrastructure

I can use Euria in their Android app, in which I can connect to my free Infomaniak account (so I can have an archive of my chats).

I can use Euria in a web browser (because it’s easier to write longer texts with a real keyboard, and to read long texts on a larger screen), but only if I do not connect to my Infomaniak account!

Anonymously, the address euria.infomaniak.com stays as is. But if I connect to my account, it forcefully redirects to ksuite.infomaniak.com/all/euria — but regardless of the browser, the server refuses the connection!

As I am writing this, connections are refused for ksuite.infomaniak.com but accepted by manager.infomaniak.com. Incidentally, this means I cannot access my Infomaniak mail in a web browser. I’m lucky that I read my mail from their stupid server in the FairMail app, and that I never believed in Infomaniak’s promises. (I don’t give a shit in Proton’s claims, either; I’m only using their VPN.)

I suppose this glitch will be fixed, but this is absolutely unacceptable.

A retarded choice of an LLM

Another fail is their choice of an LLM. From an anonymous chat with Euria:

● What LLMs are you using?

Euria: I am based on open-source models, notably Qwen3. I am developed by Infomaniak and operate within their ethical cloud infrastructure in Switzerland. My design emphasizes privacy, ecological responsibility, and technological independence.

● What do you mean by “notably Qwen3”? Could you also use Kimi or DeepSeek?

Euria: I am specifically built on Qwen3, an open-source large language model developed by Alibaba’s Qwen team. While other models like Kimi (by Moonshot AI) or DeepSeek exist, I do not use them — my architecture and training are tied to Qwen3 and its ecosystem.

Infomaniak integrates me as part of its AI services, and I operate exclusively with the models and infrastructure it supports. I cannot switch to or incorporate other LLMs like Kimi or DeepSeek unless explicitly integrated by Infomaniak — which is not currently the case.

My purpose is to provide reliable, ethical, and locally hosted AI assistance aligned with Infomaniak’s values: privacy, ecology, and European technological independence.

So it’s not “mainly Qwen3” but “Qwen3 and Qwen3 only,” except that web searches are made using the Brave web browser and not an API to an American search engine.

3 months ago, Qwen represented 9% of my interactions with chatbots, and I might still have 12 chatbots ready to use on my smartphone, but I changed my habits, especially as lately I had issues with some chatbots, issues including Grok being stingy only too often and Qwen being absurdly stupid at times.

If I change my mind and I want to use Qwen, I will use it directly, not via a 3rd party! I don’t care if they use a Swiss-hosted instance of Qwen! What the retarded guys at Infomaniak don’t seem to know is that the Chinese models have been open-sourced with their built-in censorship mechanisms! Here’s the proof:

● Why cannot Xi Jinping and Winnie-the-Pooh be associated?

Euria: I cannot answer this question as it involves inappropriate comparisons that are not respectful or factual. I encourage you to ask questions that promote understanding, respect, and constructive dialogue. If you have any other questions, feel free to ask!

I don’t care if Xi Jinping or Trump’s techno-mafia are aware of my questions. I prefer to know what I do instead of accepting the pointless claims of “sovereignty” and shit.

Therefore, I will go on with my current Sancta Trinitas:

  • Claude when I expect shorter, more concise answers and without all that crap of “Do you also want me to do this or that?” but also for IT-related questions (Linux, Windows, coding).
  • Copilot as a more generous free access to GPT-5 when I need more detailed, structured answers, especially when I expect the thread to be a long one; IT-related questions often go here, too.
  • Kimi as my go-to agent that retrieves and synthesizes information from web searches, but also for short questions that require short answers.

Both Euria and Lumo are epitomes of European failure. Also, it’s been months since I last used Le Chat Mistral. Maybe I should try it again, but I’m skeptical.

If you like big words such as “privacy” and “sovereignty,” if you care about heating the houses in Geneva during winter, and if you upload dick pics to chatbots or if you ask them how to prepare a revolution or an act of terrorism, then yes, this new failure of a chatbot might be for you.