0 models across 0 providers. The in, out and cached percentages set how much of each token type counts against API key limits.