Aleph Alpha Kolibri: How the sovereign German LLM works

driverdan | 12 days ago

IMO the announcement is better https://aleph-alpha.com/en/blog/kolibri-has-landed-a-soverei...

rpdillon | 12 days ago

Agree, and both are on the front page. Also, Show HN is usually for your own project, but this appears to be a blog post about someone else's project (albeit "a good friend").

dang | 12 days ago

It's here: https://news.ycombinator.com/item?id=49942706. I'm currently merging the threads.

pu_pe | 12 days ago

A blog post is not a Show HN topic.

embedding-shape | 12 days ago

> the weights are on Hugging Face

> Show HN is for something you've made that other people can play with

pu_pe | 12 days ago

The author did not make this model though, they used Claude to write a blog post about it.

rolymath | 12 days ago

This is not a Show HN

wingman-jr | 12 days ago

While it's not perhaps clear to me that this is a true Show HN, I enjoyed the writeup and it's good to see our German colleagues across the pond taking a good shot at this. I also appreciated the brief description on the Merlin-Arthur protocol - seems like a clever way to try to tackle the "I don't know" problem.

peterBlue75 | 12 days ago

The thing to note here, besides the transparency and the fact that it’s actually a good model that also works well on coding and agentic tasks, is that it’s the first release by a team formed less than a year ago, with a strong focus on iteration velocity. There’s more to come.

disclaimer: I‘m part of the training team, happy to answer any questions

dang | 12 days ago

Please don't copy-paste comments - it makes merging threads a pain! (and indeed the temptation to copy-paste is indication that a discussion needs merging)

p.s. It's a great comment! no problem on that level

dang | 12 days ago

We've merged (most of) the comments into this thread, which is currently on the frontpage:

Kolibri: A Sovereign Open-Weight Model - https://news.ycombinator.com/item?id=49942706 - Oct 2026 (46 comments)

Ey7NFZ3P0nzAe | 12 days ago

I'm always wondering why new models don't always adopt deepseek's KV tweaks. It's insanely valuable to have such powerful prefix caching and so cheap at inference time.

Are there drawbacks to this?