Consider the following 4 models:

Model 1 has 100 variables and 200 constraints, produces an exact solution

Model 2 has 50 variables and 50 constraints, produces a solution within 1% from the true optimal

Model 3 has 30 variables and 30 constrains, produces an exact solution

Model 4 has 5 variables and 3 constraints, produces a solution within 2.5% from the true optimal

Which of them is most appropriate? (assume that the cost of obtaining the solution grows exponentially with number of variables and constraints)