One place where I feel gemini models lag is function calling and predicting correct arguments to function calls, is there a benchmark which scores models on the

One place where I feel gemini models lag is function calling and predicting correct arguments to function calls, is there a benchmark which scores models on the basis of this ?