Consume Product

Open AI can’t go public because of how they have been spending their money. They need to bury all of that first. What they are all scared of are free models that run locally. You can now run even the largest models on PCs with low GPU and system RAM at over 30 tokens. Which makes it real time. It turns out you don’t need the whole thing in VRAM since only a small part of the model is operated on when getting a solution. By breaking the model up in ‘words’ instead of ‘letters’ and focusing on the ‘experts’ that are needed for the task, you don’t need to brute force it.

Soon local agents like Hermes will do this automatically and all you have to do is download the model on your SSD. As my Gen X nephew likes to say, “Game over, man!”