The sorting problem is one of the most fundamental problems in computer science. This paper is concerned with a new Tri-merge sorting algorithm. This is a modification of Merge sort. It is competitive with the fastest shorting algorithms, especially when the number of elements to be sorted is too large. Compared with the preceding Merge sort, Tri-merge sort is more robust. It is not only faster on random inputs, but also avoids extreme comparisons. The empirical results show that Tri-merge sort is faster than Merge sort. This reduces the time complexity and makes the algorithm faster. For a large data, we try to implement this algorithm in well known programming language Java.