Embed Notice

We all know "OpenAI" has nothing open about it, but even public models like #Llama or #DeepSeek are not open-source either despite being called that all the time.

The model weights we get are effectively a binary file resulting from the training process.

Without access to the training code we cannot train (compile) the model ourselves.
Not to mention training datasets.

All we have is the final binary.

Here's an exception though:
https://github.com/s-JoL/Open-Llama

#AI #OpenSource #DeepSeek