On August 21, 2026, AWS announced the general availability of AWS Glue 6.0, a release of its managed data integration and processing service that offers prices 30% lower than previous AWS Glue versions. The release is based on a fully updated runtime environment that includes Apache Spark 4.1, Python 3.12, and Scala 2.13, with improvements targeting the performance of extract, transform, and load jobs, as well as real-time use cases.
Broader Support for Apache Iceberg Data
AWS Glue 6.0 provides full support for the Apache Iceberg v3 specification, based in this release on Iceberg 1.11.0. Key additions include the VARIANT data type with shredding support, enabling JSON data, records, and event data to be stored and queried without flattening schemas in advance.
In practice, this approach helps data teams reduce duplicate data copies and write custom analysis code, while also reducing the likelihood of processing pipelines failing when schemas change. This is increasingly important in environments that handle semi-structured data whose structure evolves over time.
- Geometry and Geography types for processing spatial data natively in geographic information system analytics and location intelligence.
- Nanosecond-precision timestamps to meet the requirements of Internet of Things sensor data, scientific computing, and some high-frequency financial workloads.
- Unknown-type handling, enabling processing of unexpected or changing schemas without failing the data pipeline.
Spark 4.1 Engine Updates
The release includes new capabilities built on Spark 4.1 to simplify the development of ETL jobs and improve their execution. Spark Declarative Pipelines enable data engineers to describe the desired shape of transformations instead of manually managing execution order and optimization; the engine determines the execution order and appropriate optimizations.
AWS Glue 6.0 also adds native execution through Apache Arrow for Python user-defined functions, including UDFs and UDTFs. This removes the serialization overhead between Python and the JVM, which may improve PySpark performance in complex transformations.
For stateless streaming use cases, the release provides a real-time streaming mode with single-digit-millisecond latency. This mode is based on Spark 4.1’s Real-Time Mode with an optimized implementation from Glue, targeting event processing, data transformation, and routing when response time is critical.
What Changes for Users in Practice?
Taking advantage of Glue 6.0 does not require API changes. The version can be selected using the --glue-version parameter in the create-job and update-job interfaces through the AWS CLI, AWS SDK, AWS Glue Studio, Amazon SageMaker Unified Studio, and preferred development environments.
In AWS Glue Studio, new jobs can be created using the Glue 6.0 option, which supports Spark 4.1, Scala 2, and Python 3, or existing jobs can be upgraded. The Spark upgrade agent or the automatic upgrade feature can also be used, while Jupyter notebooks and interactive sessions allow version 6.0 to be specified through the %glue_version directive.
Availability and Cost
AWS Glue 6.0 is now generally available in all AWS Regions where the service operates, with regional availability varying by Region. Crawler and ETL job charges are calculated hourly and by the second, while AWS Glue Data Catalog charges are applied monthly for metadata storage and access. Storage of the first one million objects and the first one million access operations remains free, according to the information provided in the announcement.