8 pointsby hibijibies3 hours ago2 comments

ohadkran hour ago
Impressive timeline, but I’m curious whether the real advantage is model quality or simply owning the full stack and data.
- hibijibiesan hour ago
  I would think it's the fact that there's no secret sauce except data and compute to get to a decent model along with a good team. It's a Nemtron 3 Nano architecture as far as I understand, and of course there seems to be some special data curation for German.
davidos3 hours ago
one of the authors here, AMA.
- fragmede3 hours ago
  What sort of regulations did you run into while gathering training data?
  - davidosan hour ago
    The main constraint was licensing rather than regulations like GDPR. To ensure compliance, we restricted ourselves to datasets with permissive licenses.
    The one exception was the Genios dataset (a high-quality German corpus), which we used during pretraining under a separate licensing agreement. Beyond that, the growing availability of high-quality permissive datasets meant we could assemble sufficient training data without compromising on quality.