Accurate pathloss (PL) modeling is critical for 4G-LTE network planning in complex urban environments like Central Tarakan, Indonesia. This study presents a Python-based, open-source implementation of Particle Swarm Optimization (PSO) to calibrate three conventional PL models, Okumura-Hata, SUI, and Ericsson 9999, using real drive-test data. Initial RMSE values exceeded 50 dB, revealing severe inaccuracies under heterogeneous terrain. PSO optimization dramatically improved accuracy: RMSE reduced to 5.98 dB (Okumura-Hata, 89.44% improvement), 9.83 dB (SUI, 84.03%), and 6.44 dB (Ericsson 9999, 91.32%). The optimized Okumura-Hata model achieved the highest reliability, with 88.89% of measurement points meeting the <8 dB threshold and the lowest standard deviation (1.71 dB). Ericsson 9999 attained the lowest minimum RMSE (0.06 dB), showcasing exceptional potential under favorable conditions. PSO converged rapidly within 50 iterations, and sensitivity analysis confirmed that standard parameters (ω = 0.5–0.7, c₁ = c₂ = 1.8–2.2) suffice for robust calibration, eliminating need for fine-tuning. Results demonstrate that real-world propagation deviates significantly from classical logarithmic assumptions, validating the necessity of data-driven, site-specific optimization. The fully open-source framework—built with NumPy, Pandas, and Matplotlib—offers a practical, scalable solution for intelligent radio planning in dynamic urban landscapes.