JSON repeats every field name in every record and leaves types to guesswork. Avro 129 , born in the Hadoop 129 project in 2009, writes a JSON schema once and binary values only. It is the usual record format for Kafka 129 with a schema registry (Apache Kafka and Managed Cloud Kafka). The current release is Avro 1.12.2 (12 August 2026, Apache License 2.0; github.com/apache/avro (https://github.com/apache/avro 3,305 )). The examples use fastavro 1.12.2 (MIT, pip 21,050 install fastavro), a Cython library whose README reads 10,000 records in 1.7 s against 14 s for the pure-Python reference package.