Am I missing something? Llama 2 is not Open Source.

Honestly, I'm not even sure that Open Source applies to an LLM, it seems like it would be better released under an Open database license? That assumes that copyright applies to LLM weights in the first place, which as far as I know has not been tested in a court.

But regardless, Llama 2 is not Libre, it's a proprietary license that happens to allow for commercial work. I'm not sure if the article is just conflating Open Source and source available (which again... doesn't actually seem all that applicable to model weights anyway), or if the article means something else entirely when it uses the term.

They are probably just repeating facebook's ridiculous claim that it is open source. The header of the release announcement is "The next generation of our open source large language model"

As someone who has spent 25 years as an open source developer and in open source governance, I really find the abuse of the term very offensive.

Especially in this piece, where it's used as heroic embellishment.

I hope the community won't let them have this one.

[dupe]

Discussion here https://news.ycombinator.com/item?id=36774627

this is not a duplicate - it's not the announcement post, it's a deep dive into the paper from the POV of Hugging Face's RL lead.