I'm looking forward to an Agent. I have some customers who don't want US LLMs anywhere near there IP (code), but Lumo, being European and privacy focused, may convince them. Still would prefer local of course. I'm also eyeballing Vibe-cli (yes that is the actual honest to god name of Mistal's coding agent) as a "local" (as in EU) option.

What would a special Lumo agent bring to the table that you can't already get by using OpenCode with an European inference provider?

A European. Not “an”.

I'm glad that Lumo has image editing capabilities now, if it's half as good as ChatGPT, that means I no longer have to upload photos of my room to OpenAI to try a new paint color.

However, I trust Lumo only marginally more than ChatGPT, they really don't do themselves any favours by claiming it's all open-source when only the frontend is

I like Lumo, glad to pay about $12/month for a private LLM with good conversation management.

i have been underwhelmed with lumo so far. it feels a few iterations behind the current models. here's hoping its finally starting to catch up.

Proton should be receiving billions in investments from European companies, just like Anthropic and OpenAI do.

No, please. VC money gets wasted a lot, and they’ll have to make up for it by increasing prices.

Wish they'd be a bit more open source about this, bet people would contribute more.

Also about privacy-first apps.

Don't be fooled by the fact they say it's "open" on that page: only their clients are open. The models themselves? Well, good luck with that...

https://osai-index.eu/news/lumo-proton-least-open/

So they used to be just reselling access to «"Nemo, OpenHands 32B, OLMO 2 32B, and Mistral Small 3". OpenHands is a QWEN fine-tune, and Nemo and Mistral Small are both Mistral models.»

But seems like they stripped this info from their Privacy Policy now?

> But seems like they stripped this info from their Privacy Policy now?

> https://proton.me/support/lumo-privacy

"The models we’re using currently are Qwen 3.5, GLM 5.2, Image-Turbo, and FireRed-Image-Edit-1.1"

I'm not sure how old this page is, perhaps it's pre-Lumo 2.0

GLM 5.2 and Qwen 3.5 match the benchmark scores this article says their models get, so that is probably up to date.

Wow, if that is the Mark Dingemanse I think it is, he's an absolute legend in ideophone research. I used a lot of his work for my MA thesis. Always a pleasure to read.