In this paper, we propose a dynamic hybrid quantizer to reduce the burden of communication networks and on this basis, the leader-following consensus problem of first-order multi-agent systems is solved. This dynamic hybrid quantizer not only overcome the two disadvantages of the static quantizer: fixed quantization interval and infinite quantization level, but also reduce quantization error compared with dynamic uniform quantizer. Therefore, it can make the acceptable quantization errors for better system performance. Finally, the effectiveness of our approach is illustrated by theoretical derivation and numerical examples.