3 ms·

So this means llm trained on libgen and ... which is huge. Current models are only trained on "open" texts.