>>50884015I still think he probably doesn't use a model that is over 2 billion parameters max
probably like 900 or 700 million, but this is just speculation based on the way Neuro behaves compared to other similarly sized models, it could be more or less
Also I am taking a conservative number approach, as I am hesitant to assume he would be using a model that is in the hundreds of billions of parameters as that would be expensive to run and maintain locally