the context window matters more than the model name. I’ve started making this argument to clients and getting pushback. can anyone help me make the case more precisely or tell me where the argument breaks down?
the argument holds for task-specific use cases. if your task is ‘process this 50-page document,’ the model that can hold the whole document in context outperforms a better model that can only hold 10 pages, regardless of other capabilities. context window is a hard constraint. model quality is a soft one.