<p>Operating speed is a critical parameter for assessing safety and geometric design consistency (GDC), especially on two-lane mountainous highways. This study aims to develop vehicle-specific operating speed prediction (V₈₅^P) models that estimate the 85th percentile speed to evaluate GDC for the safety assessment of horizontal curves. Continuous speed data were collected using a GPS device for three vehicle categories (two-wheelers, passenger cars, and trucks) across 48 horizontal curves. Traditional regression approaches demonstrate weak performance owing to geometric complexity and traffic heterogeneity, necessitating curve classification and the backward stepwise regression (BSR) technique. The refined models developed in the present study achieved strong predictive accuracy with R² values ranging from 0.755 to 0.856. GDC evaluation using operating speed safety criteria revealed that 65% of the curves were rated Fair or Poor, particularly those with design speeds below 30&#xa0;km/h. These findings underscore the influence of vehicle type and curve classification on the operating speed. The developed models can effectively assess safety in mountainous terrain using GDC and recommend targeted interventions such as traffic warning systems, speed management strategies, and geometric design improvements. These models serve as practical decision-support tools, enabling speed simulations and informing design revisions during both the planning and post-construction phases. A field study underscores the applicability of these models as a reliable tool for highway authorities and engineers to optimize highway geometric design and enhance safety.</p> Graphical abstract <p></p>