Quarkus has great performance – and we have new evidence

The Quarkus team published a new transparent, reproducible benchmark comparing Quarkus and Spring Boot performance. Results show Quarkus handles 2.7x more transactions per second (19,255 vs 7,238 tps), starts 2.3x faster, and uses half the memory. The benchmark addresses past shortcomings: outdated data, missing throughput metrics, and lack of reproducibility. The team open-sourced the benchmark code, invited Spring Boot community input to ensure fairness, and explored questions like virtual threads impact (+6k tps for all frameworks) and Spring Boot 3 vs 4 differences. The post also clarifies that while Quarkus JVM mode outperforms alternatives across all metrics, native compilation does cut throughput in half (though startup and memory improve dramatically), making native mode best suited for frequently restarted or low-workload applications.