The classification of cancer based on gene expression data is one of the most important procedures in bioinformatics. Classification of leukemia types still remain as an unsolved problem in the medical history despite many years of efforts. Micro array datasets for Leukemia were pooled from several databases and classified using supervised and unsupervised learning techniques. Support Vector Machines has been used to solve multi-class classification of leukemia datasets. Linear Kernel was found perform best in the classification of Leukemia microarray datas as reported by Golub et al. Performance of Linear SVM was evaluated using the prediction accuracy explained by Baldi et al. SVM classified the three leukemia classes with overall accuracy of 83.3% and GC2 value of 0.73%. Experiments performed on gene expression datasets demonstrate that SVMs are known to outperform all the other supervised classification algorithms. In this post genomic era, SVM approach can enhance the successful classification of leukemia datasets during clinical trials of leukemia cases.