The debate on whether intelligence is influenced by or not in grouping gifted students for leaning necessitates this study. This paper therefore, established empirically a position on levels of intelligence in cooperative learning for Mathematics achievement of gifted students. The study is based on a pretest-posttest quasi-experiment. Sixty-one participants were selected through multistage screening and were randomly assigned to experimental group and control. The experimental group was exposed to cooperative learning, while control with conventional learning in Mathematics. Twelve weeks period were used for the study. Instruments used were Mathematics Achievement test (r=0.89), school academic record; nomination checklist and adapted Slosson Intelligence Test (reliability=0.89). The data collected were analyzed using ANCOVA and MCA. Findings revealed that cooperative learning and levels of intelligence have significant interaction effect on mathematics achievement. Grouping of students for cooperative learning was revealed to be best at ratio 1:2:1 for high, superior and very superior intelligent students respectively.