Top
Best
New

Posted by denysvitali 9/2/2025

Apertus 70B: Truly Open - Swiss LLM by ETH, EPFL and CSCS(huggingface.co)
322 points | 61 comments
kriro 9/5/2025|
Really happy to see this and will give it a good spin. They seem to be doing things the right way in my subjective opinion:

""" To implement this filter, we begin by ranking URL domains according to the volume of texts they contribute to the FineWeb (Penedo et al., 2024a) and FineWeb-2 (Penedo et al., 2025) corpus, as an approximation of web-level English and multilingual data. From this ranking, we select the top one million English domains and the top one million non-English domains. Due to domain overlap and the fact that some sites are now offline, the total number of accessible robots.txt files is smaller than two million. For each domain that remains reachable, we retrieve its robots.txt file as of January 2025 and examine the directives relevant to AI training. In particular, we focus on those targeting the AI-specific user agents listed in Appendix A. Any contents blocked by the current robots.txt is removed retroactively from the entire 2013-2024 range of the training dataset. We follow an opt-out policy, that is, if the corresponding robots.txt files are not available, we consider the data usable for training. The filtering process results in an estimated token loss of approximately 8% in English data and 4% in multilingual data. """

mycall 7 days ago|
> Any contents blocked by the current robots.txt is removed retroactively from the entire 2013-2024 range of the training dataset

Why not check historical versions of the robots.txt (e.g. archive.org) and contain the retroactive cutoff to a certain date range, parsing the robots.txt accordingly? That might increase the corpus size within legal and fair use boundaries.

lllllm 7 days ago|||
common crawl anyway respects the CCbot opt-out every time they do a crawl.

we went a step further because back in old ages (2013 is our oldest training data) LLMs did not exist, so website owners opting out today of AI crawlers might like the option to also remove their past contents.

arguments can be made either way but we tried to remain on the cautious side at this point.

we also wrote a paper on how this additional removal affects downstream performance of the LLM https://arxiv.org/abs/2504.06219 (it does so surprisingly little)

pdpi 7 days ago|||
"I didn't know to withdraw consent" isn't the same as "I consent". Thank you for doing the right thing.
mycall 7 days ago|||
Ah good points, thanks for the clarification.
3np 7 days ago|||
I imagine coverage is sparse enough to not be worth it.
denysvitali 9/2/2025||
Report: https://github.com/swiss-ai/apertus-tech-report/raw/refs/hea...

Key features

Fully open model: open weights + open data + full training details including all data and training recipes

Massively Multilingual: 1811 natively supported languages

Compliant: Apertus is trained while respecting opt-out consent of data owners (even retrospectivey), and avoiding memorization of training data

lyu07282 9/3/2025||
Their struggle with Nvidia driver bugs they had to work around was very relatable. You'd think if someone buys 10,752 of their high-end GPUs you'd get some support with it.
hodgehog11 7 days ago|||
Agreed, but the problem seems to be even worse with AMD from what I hear, or at least it was when I checked with some of my HPC buddies a little over a year ago. Constant driver bugs and crickets from upstream "support".
_zoltan_ 9/5/2025||||
did I miss a blog on this?
lllllm 9/5/2025||
we didn't have time to write one yet, but there is the tech report which has a lot of details already
menaerus 7 days ago||
Report is packed with interesting details. Engineering challenges and solutions chapter especially show how things which are supposed and expected to work break when put through a massive scale. Really difficult bugs. Great writeup.
lllllm 7 days ago||
thank you!
hhh 7 days ago|||
no, you have to pay the yearly per gpu license for that.
Bromeo 9/2/2025|||
Looks like the performance is pretty decent, somewhere around Llama3.1 for general knowledge (Tables 17) but still a bit behind in Code and Reasoning (Table 18). Llama3.1 was released about one year ago.
esafak 9/5/2025||
There's an interesting "Swiss AI Charter" on pg. 107.
lllllm 7 days ago||
martin here from the apertus team, happy to answer any questions if i can.

the full collection of models is here: https://huggingface.co/collections/swiss-ai/apertus-llm-68b6...

PS: you can run this locally on your mac with this one-liner:

pip install mlx-lm

mlx_lm.generate --model mlx-community/Apertus-8B-Instruct-2509-8bit --prompt "who are you?"

trickstra 7 days ago||
Hi, your "truly open" model is "gated" on Huggingface, restricting downloads unless we agree to "hold you harmless" and share our contact info. Can you fix this please, either by removing the restriction, or removing the "truly open" claim?
lllllm 7 days ago||
We hear you, nevertheless this is one of the very few open-weights and open-data LLMs, and the license is still very permissive (compare for example to Llama). Personally of course I'd like to remove the additional click, but the universities also have a say in this.
dougnd 7 days ago|||
This project looks awesome!

In the US, many state governments have anti-indemnify laws that restrict the state government agencies (including state universities) from agreeing to contracts and agreements with such language. I'd love to make this available to researchers at my university, but I'm not sure I can click through such an agreement (similar problems exist with other LLMs).

It is Apache 2 and I don't see anything that prohibits another contracting party from agreeing to the Apertus LLM Acceptable Use Policy and redistributing with just Apache 2 and without the AUP. Maybe this provides a solution? Unless I'm missing something?

lllllm 7 days ago||
yes this seems a good way to go. for example you can already find many quantized versions under https://huggingface.co/models?search=apertus%20mlx and elsewhere
trickstra 2 days ago|||
Ok so why keep calling it "truly open" then? It's an obvious lie and nobody is forcing you to say it. It benefits your marketing, sure, but it harms everyone else by diluting the meaning of the term "open". So stop doing that please.
trcf22 7 days ago||
Great job! Would it be possible to know what was the cost of training such a model?
menaerus 7 days ago||
From their report:

> Once a production environment has been set up, we estimate that the model can be realistically trained in approximately 90 days on 4096 GPUs, accounting for overheads. If we assume 560 W power usage per Grace-Hopper module in this period, below the set power limit of 660 W, we can estimate 5 GWh power usage for the compute of the pretraining run.

nickpsecurity 9/2/2025||
Upvoting to encourage discussion of these differentiators:

"Apertus is a 70B and 8B parameter language model designed to push the boundaries of fully-open multilingual and transparent models. The model supports over 1000 languages and long context, it uses only fully compliant and open training data, and achieves comparable performance to models trained behind closed doors."

"pretrained on 15T tokens with a staged curriculum of web, code and math data"

"open weights + open data + full training details including all data and training recipes"

"Apertus is trained while respecting opt-out consent of data owners (even retrospectivey), and avoiding memorization of training data"

Mars008 9/3/2025|
At least not "open source"

> "open weights + open data + full training details including all data and training recipes"

Is it reproducible?

> respecting opt-out consent of data owners (even retrospectivey)

Were they notified and given an option to opt out? Owners and authors are not the same. Data owners aren't copyright owners either.

> avoiding memorization of training data

Not convincing.

ujjkel9938 9/3/2025||
I saw some of the pretraining code in github, but not the post-training.
lllllm 9/5/2025||
posttraining codebase is here: https://github.com/swiss-ai/posttraining
lastdong 9/2/2025||
In my opinion, we need more models trained on fully traceable and clean data instead of closed models that we later find out were trained on Reddit and Facebook discussion threads.
johntash 7 days ago|
I want to see something trained _only_ on stuff like encyclopedias, programming books, etc. I'm interested in how different it would be compared to something with a lot of social media in it.
ekianjo 7 days ago||
Better to do a fine tune or a LoRA than a full retraining from scratch
dkga 7 days ago||
Congratulations to the Apertus team! Love the name, which in addition to "open" in Latin reminds me of Pilatus.
djoldman 7 days ago||
This is why predictions of eventual AI failure due to copyright lawsuits are likely wrong.

Assumedly, an organization training and then distributing this model cannot be stopped via copyright or breach of contract lawsuit. It may be that folks will figure out copyright-free versions of text-to-image and text-to-video, etc., models as well.

It seems that there is plenty of copyright-free data available to train a useful model. Therefore, when content creators upset about AI companies training models on their content are asked about this model, they have nothing to do but shrug.

The cat is out of the bag.

dcreater 9/5/2025||
I want and hope this to succeed. But the tea leaves don't look good at the moment:

- model sizes that the industry was at 2-3 gens ago (llama 3.1 era) - Conspicuous lack of benchmark results in announcements - not on openrouter, no ggufs as yet

lllllm 9/5/2025|
benchmarks: we provide plenty in the over 100 page tech report here https://github.com/swiss-ai/apertus-tech-report/blob/main/Ap...

quantizations: available now in MLX https://github.com/ml-explore/mlx-lm (gguf coming soon, not trivial due to new architecture)

model sizes: still many good dense models today lie in the range between our small and large chosen sizes

dcreater 9/5/2025||
Thank you! Why are the comparisons to llama3.1 era models?
lllllm 7 days ago||
we compared to GPT-OSS-20B, Llama 4, Qwen 3, among many others. Which models do you think are missing, among open weights and fully-open models?

Note that we have a specific focus on multilinguality (over 1000 languages supported), not only on english

kamranjon 7 days ago|||
How did it compare with Gemma 3 models? I’ve been impressed with Gemma 27b - but I try out local models frequently and I’m excited to boot up your 70b model on my 128gb MacBook Pro when I get home!
dcreater 7 days ago|||
ah im sorry, I missed that - im not that blind usually..
SilverElfin 9/2/2025||
Apparently a project of https://www.swiss-ai.org/
WanderPanda 9/5/2025|
Imagine regulators doing their job for once and creating a clean regulation that removes the uncertainty about the liability for such releases. Such that they can just slap Apache or MIT on it and call it a day and don't require to collect personal data to comply with the "acceptable use policy".
More comments...