How We Made Ktor JSON Parsing 66% Faster with 50% Less Memory on KMP

If you are building Kotlin Multiplatform applications, you know that JSON serialization is often one of the main bottlenecks in high-throughput backends and mobile apps—especially regarding CPU cycles and Garbage Collector (GC) pressure.

I’ve been working on Ghost Serializer—a byte-first JSON serializer designed for KMP, and the official HTTP Arena benchmarks just went live showing some wild improvements when integrated with Ktor.

Don't take my word for it. Here is the raw comparison of Ktor base vs. Ktor + Ghost under load:

JSON Throughput: Slashed latency to jump from 395k to 658k rps (+66.6% speedup).

Heap Memory Consumption: Plunged from 9.90 GiB to 4.80 GiB (Over 50% memory reduction).