In zero-temperature QCD it is generally BPST instantons. These are
known to explain spontanteous breaking of $SU(N_f)_A$ chiral
symmetry, which provides quarks with constituent mass. These
gauge-configurations alone do not give rise to (i.e. "explain")
confinement.
In non-zero termperature QCD with vanishing holonomy, it is KvBLL
calorons.
With non-zero temperature and non-zero holonomy, it is the and
instanton-dyons. These are suspected to possibly explain confinement.
The semi-classical approximation can be applied to all of these dominant configurations, which basically means approximating the path integral about these configurations via the saddle-point approximation. Despite the fact that this approximation integrates over the dominant contributions to the path integral, higher-order perturbative fluctations can (and usually do) play a crucial supplementary role (e.g. solving the large instanton problem). Their effects and often are implemented phenomenologically.
You can read about all this information in this lecture series by Edward Shuryak. [arxiv:1812.01509]