Not really. Local models are pretty decent for simple tasks. The hardware to run them costs less than a month’s salary.