Stochastic fatigue crack growth (FCG) modeling is vital for fatigue reliability and durability analyses of engineering components. To characterize the statistical properties of FCG process, the Yang–Manning model was introduced to calculate the crack exceedance probability and fatigue life distribution to reach any crack size; however, it is only applicable for failure probability analyses of pre-cracking specimens. According to this, a new probabilistic fatigue life model considering the distribution of initial crack sizes was proposed. Experimental data of Al 2024-T3 alloy and 30NiCrMoV12 steel were utilized for model validation and comparison. Results show that the proposed probabilistic model works well for FCG analysis with independent cracks. In addition, the Yang–Manning model was extended to consider the crack coalescence issue under multiple cracks condition, which shows a sound conformity to experimental data.