In the market for a high-performance, cost-efficient database for your AI-powered apps? Vector search is just one of many critical areas where Couchbase’s unified data platform measurably beats MongoDB™. This new vector search benchmark report compares Couchbase 8.0 and MongoDB 8.0 for recall, latency, and throughput using VectorDBBench, the standard benchmarking suite for vector workloads. Couchbase significantly outperformed MongoDB for both standard and large-scale datasets:
- 100M vector dataset with 768 dimensions – Couchbase achieved higher throughput for all recall levels while providing higher levels and lower latencies
- 1B vector dataset with 128 dimensions – Couchbase’s performance degraded only slightly, while MongoDB could no longer achieve any production-grade performance
The report includes full details of the testing methodology, hardware setups, and workloads, plus a discussion of the results.