NHacker Next
- new
- past
- show
- ask
- show
- jobs
- submit
login
self.__VINEXT_RSC_CHUNKS__=self.__VINEXT_RSC_CHUNKS__||[];self.__VINEXT_RSC_CHUNKS__.push("2:I[\"aadde9aaef29\",[],\"default\",1]\n3:I[\"6e873226e03b\",[],\"Children\",1]\n5:I[\"bc2946a341c8\",[],\"LayoutSegmentProvider\",1]\n6:I[\"6e873226e03b\",[],\"Slot\",1]\n7:I[\"3506b3d116f7\",[],\"ErrorBoundary\",1]\n8:I[\"a9bbde40cf2d\",[],\"default\",1]\n9:I[\"3506b3d116f7\",[],\"NotFoundBoundary\",1]\na:\"$Sreact.suspense\"\n:HL[\"/assets/index-BLEkI_5r.css\",\"style\"]\n")lass="_meta_ka9gd_33">133 points by muzzy19 2 days ago | 69 comments
Rendered at 10:56:59 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
Since all frontier models are owned by US companies, I think better alternative is to focus on open source models only that run on EU data centers owned by EU companies. That will be something.
And, more importantly, that you can use the existing OpenAI SDK for your language but swap models (even across providers) by changing one line of code.
You're paying for convenience, yes, but model routers solve a real problem.
"allegedly"
Worth mentioning that Huggingface already offers a similar service. And they are also European:
https://huggingface.co/docs/inference-providers/index
https://huggingface.co/inference/models
Very ill-suited comparison to IBM.
It's well buried though. Does not seem to be a focus of theirs.
Edit: looking more closely, it's the company orginally known as "Datagenius SAS" based in Lyon, France. They changed their name to "Eden AI SAS" in 2022 (or maybe it's an alternative name? I am not too familiar with how this works in France), and their datagenius homepage links to the submitted page. https://www.datagenius.fr/ If anyone wants to send them a letter, the registered address of the company is 142 Rue De Crequi, F-69003 Lyon
That took slightly more work to figure out than I would expect from a website that has the word "transparency" in the headline, but at least they do exist
Not sure why you'd sound so confident and not qualifying it somehow when it seems you don't actually know what you're talking about, and it's so easy to lookup before spewing wrong information.
So 100% with you here...
If Eden provides a similar feature set, I'd certainly consider them.
In other words, it solves an organizational problem, not a technical one. That’s what the 5.5% is for.
Whether or not you prefer this or OpenRouter or one of the other LLM gateways is another discussion.
But the most important advantage is the convenience of being able to try out new models without subscribing to yet another service.
Edit: Somehow, the comment got deleted although I replied to it? It originally implied I wouldn't be able to explain why it's easier, so I wrote the above to explain.
No, it's not. You're absolutely right on that.
It's just someone you don't know who actually runs it due to no proper imprint promoting their business over someone else who you also don't know who actually runs it. So you send all your valuable business data to unknown guy A instead of unknown guy B. Oh, and also, in both cases you couldn't even sign a proper data subprocessing agreement with both guys. You can't sign it with guy A, who doesn't care, and you also can't sign it with guy B who says he's from Europe, does not even bother to provide an address to prove that, and obviously does not understand the GDPR.
Net souvereignty gain is zero by switching the middle man. In fact I'd say using such a "European" router service is actually worse than making business directly with, let's say, AWS, OpenAI or Anthropic where you'd at least know where you're buying from.
Probably for quite a few people.
Yes, that's why I switched
What happens after the AI Gateway don't matter that much, since the whole purpose of the product seems to be about routing LLM inference requests, if it didn't do that, I don't think they'll have anything to sell in the first place :)
"European Alternative" has a different connotation as visible in the other comments.
it could also be made in the EU and not hosted there
You can use it with just European models if you want.
I don't think that's the case with this particular company. It's not clear who's running the show, it's not clear if they abide by any of the EU regulations, and their lack of proper documentation probably makes them more of a liability than an asset. Plus, for any of it to hold water, you'd need to set up all kinds of paperwork with the people providing the compute if you don't want to be just as impractical a partner as the American competitors.
The business model is still sound.
aws, azure, cohere, mistral etc on here are all linked to generic privacy policies, it is impossible to tell what retention sla is in place.
to the credit of openrouter, they do state the data retention policies of almost all, but not all of their providers.
you can check the retention level for models on openrouter with the providers list. i was not able to find a retention policy for their plugin providers (exa and parallel). both log by default with enterprise opt in for zdr.
We don’t brand ourselves as such, but if you’re looking for a European OpenRouter alternative focusing on media models (image, video etc.), I‘ve built that with https://lumenfall.ai
- no zero data retention available for the router itself (rights to retain data flagged by content scanning).
- no stated privacy policies for the providers served on the api. quote: "we work with providers to understand their policies and can provide information about specific providers on request".
on openrouter, on the positive side, they have transparent retention agreements for almost all providers, and also offer zdr.
i wish there will be a time where privacy is for everyone, not just for people who can afford an enterprise sla
How is it sustainable?
Besides that, we‘ll introduce paid plans with additional features like observability soon.