The Byzantine Fault Tolerance State Machine replication (BFT-SMR) protocol ensures the correct operation of distributed systems in the presence of faulty or malicious replicas. Traditional Byzantine protocols typically employ a single-leader model to drive consensus, whereas this model exhibits significant performance bottleneck: (1) The single-leader model requires all replicas to participate in the consensus, resulting in increased communication overhead among replicas and higher consensus latency; (2) The single-leader model incurs a communication cost of \(O(n^2)\) during the timeout phase. To address these issues, this paper proposes a BFT-SMR protocol, GP-BFT, based on multi-committee parallel consensus. GP-BFT initially divides online replicas randomly into multiple committees, forming a committee set, and follows a 2-chain commit mechanism to achieve parallel consensus among committees. By transitioning from full-replica consensus to intra-committee consensus, GP-BFT reduces commit latency while ensuring safety and liveness. Experimental results demonstrate that in scenarios with large-scale replicas and transactions, GP-BFT maintains high throughput with lower latency.