Comment by nphardon

Is a special hardware setup required to run open weight models? It's not clear to me from the article. Also, what does paying for inference mean? I thought the inference step happened during the training of the models, and that by the time I am interfacing with an LLM all the inference has been long done.