Is Llama 3.1 truly "open-source"?

Question

Raju Sivaram · Accepted Answer

Yes, Llama 3.1 is indeed an open-source AI model. Meta has released Llama 3.1 as the largest-ever open-source AI model, which the company claims outperforms other prominent models from companies like OpenAI and Anthropic. https://www.theverge.com/2024/7/...

Rafael Pinheiro · Answer

@raju_sivarama_gadiraju got your point, but for LLMs to be truly "open-source," they need to include: 1: Detailed info about the training data so that someone can recreate a similar system with similar data. 2: The source code used to train and run the system. Llama 3.1 seems to lack those requirements. For more on the definition of Open Source AI and this debate, check out: https://opensource.org/deepdive/...

Kabir · Answer

Wihile LLaMA models are open source in the sense that their weights, code, and doc are available for download and use, there are still some limitations that prevent full reproducibility and democratization of AI. For example the training data that used is not available so you cannot use the same code to reproduce a similar model. And of course the compute and GPU resources needed to reproduce is beyond most startups and all individual tinkerers. In traditional open source projects - you download source, change a few header files and a few methods and recompile and you have your own flavor of the “distribution” be it Apache, Linux, Python, PHP, PGP.

Is Llama 3.1 truly "open-source"?

Replies