Performance Benchmark of Common Java JSON Libraries Using JMH

In this article the author evaluates the performance of four popular Java JSON libraries—Gson, Fastjson, Jackson, and Json‑lib—by writing JMH benchmarks for both serialization and deserialization of a complex Person model.

The article first introduces each library, their background, Maven dependencies, and provides simple utility classes ( FastJsonUtil , GsonUtil , JacksonUtil , JsonLibUtil ) for converting between Java objects and JSON strings.

A representative Person class with nested objects, collections, and maps is defined to simulate real‑world data. Two JMH benchmark classes ( JsonSerializeBenchmark and JsonDeserializeBenchmark ) are presented, each parameterized with different operation counts (1 000, 10 000, 100 000) and measuring single‑shot execution time.

Results, exported to ECharts charts, show that for small counts Gson is fastest, while at larger counts Jackson and Fastjson outperform Gson; Json‑lib is consistently the slowest. Deserialization performance is similar among Gson, Jackson and Fastjson, with Json‑lib again lagging.

The article concludes with a brief commentary on choosing a JSON library based on performance needs and invites readers to join the author’s community for further discussion.