Road freight transportation connects producers and consumers, and delivers goods in a timely manner, which is essential for the economy and society. However, traditional freight forwarding faces challenges such as long delays, high labor costs, and empty mileage, which increase the logistics and environmental costs and affect the carriers and shippers negatively. To address these issues, we develop a sustainable smart freight-matching model using Reinforcement Learning (RL). The main components of matching platform include: (i) optimizing real-time carrier-shipper matching using RL, aiming to maximize matches while considering time, location, and capacity (ii) emphasizing designing a dynamic dispatching system to ensure on-time availability, optimize resource allocation, enhance customer satisfaction, and improve operational efficiency in freight fleet management, (iii) stablishing dynamic cargo consolidation to reduce shipping costs, lower CO2 emissions, and enhance vehicle and logistics resource efficiency. To achieve sustainability, this platform maximizes platform profit as economic criterion, minimizing vehicle emissions, as environmental criterion and maximizing service level, as a social criterion. We use the Actor-Critic framework to model the state, action, and reward of the smart freight platform. We use Montreal region as a case study to test our model. We create synthetic data that simulates the real-world characteristics of freight logistics, such as timestamps, player roles, and location coordinates.