Anonymous functions are the same as regular functions (except they're restricted to one-liners with no branching or assignment). They accept any type of input argument of any size, scalar, vectors, arrays of numbers, structures, objects, etc. As long as the code of your function can deal with vectors, you can pass it a vector.
In your case, the anonymous function is already capable to handle vector osf since the only function that uses it, log, can:
>> fptest = @(osf, T) T^2 - log(osf);
>> fptest([1 2 3], 0)
0 -0.69315 -1.0986
Note that I would change the order of arguments, or the order of the formula in fptest, so that they match. It minimises the chance of you calling the function with the arguments reversed:
fptest = @(T, osf) T^2 - log(osf);
fptest = @(osf, T) -log(osf) + T^2;