Where’s the source code?

Did you measure the warmup time with the actual parsing time? For example, some libraries have heavy initialization but then perform very fast afterwards. It’s probably what you see with Jackson here.

Did you benchmark on an Android device or the JVM?

If you benchmarked on a device, you should seriously consider using the Android benchmark library which addresses a lot of common issues with device measurement, like the effects of thermal throttling.

And finally about Moshi: did you use Moshi’s Kotlin codegen plugin?