> OpenEuroLLM has a total budget of €37.4 million of which €20.6 million comes from the Digital Europe Programme.<p>So on average 1.87 million per participating institution which might amount to funding ~5 PhD students per institution. Not bad for a training program.<p>The project has been awarded the Sovereignity Seal, an EU mark of Excellence before it even started. This is truly in accordance with european values, where we reward participation and proclamation. I don't think we will ever hear again from this project.<p>Congratulations to the participants of the consortium for receiving this large EU grant. Thoughts and prayers to the students who will be writing the deliverable progress reports.
My comment from the original submission [1]:<p>---
As someone who is in general skeptical of programs like this (and an European) there are 2 remarkable / timely things about this:
- This project doesn't just allocate money to universities or one large company, but includes top research institutions as well as startups and GPU time on supercomputing clusters. The participants are very well connected (e.g. also supported by HF, Together and the likes with European roots) - Deepseek has just shown that you probably can't beat the big labs with these resources, but you can stay sufficient close to the frontier to make a dent.<p>Europe needs to try this. Will this close the Gap to the US/China? Probably not. But it could be a catalyst for competitive Open source models and partially revitalize AI in Europe. let's see..<p>PS: on Twitter there was a screenshot yesterday that in a new EU draft, "accelerate" was used six times. Maybe times are changing a little bit.<p>Disclaimer: Our company is part of this project, so I might be biased.
---
I hope the next time this is on HN, it's with some cool release and not a PR :).<p>(@mods please delete if copy-quoting not allowed)<p>[1] <a href="https://news.ycombinator.com/item?id=42924802">https://news.ycombinator.com/item?id=42924802</a>
It is worth noting there is _another_, completely unrelated project (also) called *EuroLLM* that is also EU funded which not only shares many of the same goals, but has already fulfilled many of them:<p>1. large multilingual dataset<p>2. open science approach<p>3. competitive performance<p>Here is the HF blogpost that introduced it in December 2024 (along with various benchmarks): <a href="https://huggingface.co/blog/eurollm-team/eurollm-9b" rel="nofollow">https://huggingface.co/blog/eurollm-team/eurollm-9b</a><p>The project's lead has summarized the situation succinctly in their LinkedIn post [0]<p><pre><code> I hope the different communities collaborate openly, share their expertise, and don't decide to reinvent the wheel every time a new project gets funded. Next what? "OpenEuroLLM with real cheese"?
</code></pre>
[0] <a href="https://www.linkedin.com/posts/andre-martins-31476745_ai-artificialintelligence-openlanguagemodels-activity-7292215869395394560-Msbl" rel="nofollow">https://www.linkedin.com/posts/andre-martins-31476745_ai-art...</a>
This should probably link to the actual press release since its more of an announcement of something forming rather than a release of any models, code, whitepapers, etc.<p><a href="https://openeurollm.eu/launch-press-release" rel="nofollow">https://openeurollm.eu/launch-press-release</a>
<i>But can I run it on Gaia-X?</i><p>This really reads like a parody. Press release, “a consortium of 20 research institutions”, “awarded the STEP (Strategic Technologies for Europe Platform) seal”. Lots of grandiose self-congratulations. All with nothing to run, download or try of course.
> A series of foundation models for transparent AI in Europe<p>Am I the only one who doesn't see any link to any model? Too many words, no actual outcome.
The three goals featured prominently above the fold are:<p>> truly open
> including data, documentation, training and testing code, and evaluation metrics; including community involvement<p>> compliant
> under EU regulations, OpenEuroLLM will provide a series of transparent and performant LLMs<p>> diverse
> for European languages and other socially and economically interesting ones, preserving linguistic and cultural diversity<p>The first one seems good, but the second two seem to be pretty beside the point of creating models that compete with the cutting edge of China and the USA.
》under EU regulations, OpenEuroLLM will provide a series of transparent and performant LLMs<p>What EU relulations? It is a moving target, and nobody knows what exactly apply. It would be nice to provide list of regulations with references. And some testing suite or checklist, to verify AI use actually fits regulations.<p>Right now, if I integrate spellchecker into my app, I have no idea if I am breaking any AI EU regulations!!!
Apart from the already mentioned pessimism what matters is if it is great when used with all European languages. It might help break some usual boundaries and allow better RAG between the resources in different countries. English speaking folks take it as granted that everyone uses it in business but that is not the case in Europe. And no US company is interested in solving that provblem.
They also had some search engine announced with similar name like "openeurosearch" or something close to that.<p>That project too seems dormant lately.<p>They just announce things and then the train leaves the station.
It's funny they don't have Hugging Face as their Partner. Literally, the biggest face of Open LLMs sitting right in Europe, but somehow it's not a partner.
How would EU enforce their regulation with an open recipe? Say I take the recipe plus my own data to train a model that has no check of my speech whatsoever, and let my friends use it, wouldn't that violate some regulations of EU?
Where is it? Where is code, a model, a repo? If one of our startups in this community put out a press release like this saying they have plans they’d be laughed off HN. Why isn’t the EU held to the same standard?
How can a project be really "truly open", "compliant", and "diverse" if it does not even have contact information on its main page. (It is hidden in the press release.)
Honestly, the cynicism in the comments here is extremely disappointing.<p>EU citizens badly need AI systems that are open and privacy-respecting. Getting together this rather large coalition of experts with quite some money and (importantly) access to compute power is a nice first step.<p>Let them play around, train some models, fail-and-get-up-again, start over, write papers and hopefully get some useful output. Remember, for the involved PhD students it will also be a learning experience!<p>Yes, it's only the first step. But yeez, it's a press release indicating the start of a scientific collaboration! Let's hold back on the negativity for a couple of years until after they've had a chance.<p>I, for one, hope this will lead to success and wish the team the best.
just smells really "EU". lots of talk about inclusion, regulations, equality, lot of kissing the ring of institutions. they are cooked aren't they? :(
I don't fully understand the reactions in the comments.
It's an announcement of a project that is starting soon-ish, mostly aimed at recruiting students and getting European talent excited, why would you expect them to deliver a model upon announcing the project?<p>Did you expect OpenAI to release GPT in the press release that announced its creation as a company?
Bullshit Silicon Valley startups do big press releases based on literally nothing all the time, but all of a sudden this is an issue if an academic European institution does it?<p>I hope the post is being bot-raided because otherwise I'll have to accept that the quality of thought on HN has gone down. I get the typical biased US-elitism that is pervasive on this website, but these reactions are just plain dumb.
Where is the models - I don't care about your "intentions" and your logos - just make and publish the models. This current form is _useless_
From the web site: Open, compliant, diverse, yada yada.<p>In real life: you either train your LLM on Anna's archive, or get left behind with sub-par model
Open but to what degree. Not saying that I will but can I generate hate speech with it. If not why not. What about other “deemed illegal by non elected officials” activities?