Assessment of reliability in the early stages of software development from architectural models is one of the major challenges that many studies have addressed in this field in the last decade. The main drawbacks of existing methods are the following: 1) considering equal impact for all parts of the software architecture on system reliability, and 2) inability to determine the contribution of each part of the software in the system failure. This paper introduces the extended version of the colored petri net as an underlying verifiable model to evaluate the reliability of a software system. The proposed model enhances reliability assessment of the software system by measuring the density of failure for each part of the software system, predicting the reliability during execution of a scenario, and estimating the reliability of every structural part of a program, such as loops and conditions. These innovations enable software architects to cost-effectively identify and correct the vulnerable parts of a system in the early stages of software development. The high-level model of the scenario is taken as a UML sequence diagram. Synthesis of the formal model is conducted using an introduced graph called a fragments dependency graph. The practicality of the suggested approach is illustrated by a case study.