In networking, Lots of routing mechanisms in the Internet network have been based on single-shortest path routing in various metrics. To transfers the data to one end to another means it heavily based on the shortest paths in different node pair. Hence, routing protocols based on single-shortest path method cause the severe convergence of packets to same links, and results to traffic congestion. If the same links are would be often used, and some alternative links become new bottlenecks. Meanwhile, the techniques of multi-path routing have been focused as another methodology to ease traffic congestion. Hence, multi-path routing encourages the effective utilization of network resources. In addition, multi-path routing means to use more links than single routing, and it is favorable in terms of the load-balancing of traffics. Consequently, multi-path routing contributes the various things, such as the improvement of the quality of service (QoS), minimizing delay, the increase of reliability of delivery, and so on. In our Paper, We present a new multipath routing methodology called MLB-routing (multinomial logit based routing). MLB-routing is the best multipath routing because it reduce the convergence in link and increase the bandwidth.MLB routing incorporate the multipath from the same origin and destination. Using request reinforcement message send from sink we can send our data to destination without any delay or loss. And also we can implement this model into based on both link state and distance vector based protocol.