In the era of Noisy Intermediate-Scale Quantum computing, effective quantum circuit compilation is critical to bridge high-level quantum logic with the constrained hardware of Quantum Processing Units. This study empirically evaluates the transpilation phase—encompassing gate decomposition, mapping, and routing—of three leading quantum circuit compilation frameworks: IBM Qiskit, Google Cirq, and Tket, targeting the IQM Spark 5, a gate-based QPU architecture previously unstudied in this context. A benchmark of 50 circuit instances, representing GHZ, Hamiltonian simulation, Mermin–Bell inequalities, and QAOA algorithms, was designed to capture diverse qubit counts, gate compositions, and structural complexities. Each instance was compiled 100 times per framework, evaluated on output circuit depth, two-qubit gate count, and compilation time. Stability was assessed via standard deviations, alongside average performance and comparative analysis by instance type. Results show Qiskit often excels in circuit quality, while Cirq offers faster transpilation for specific cases; no framework is universally superior, based on the performed experiments.