Java 25, the latest Long-Term Support (LTS) version of the Java Development Kit (JDK), is now available for download. This release packs in 18 new features, with five specifically designed to help with artificial intelligence (AI) development. It follows JDK 24, which came out just six months ago. Since this is an LTS release, Oracle promises at least eight years of support, making it a solid choice for long-term projects.

What’s New in JDK 25

One of the big upgrades is a new way to profile CPU usage on Linux. The JDK Flight Recorder (JFR), Java’s built-in tool for monitoring and profiling apps, now can better track how much CPU time Java programs use on Linux systems. This helps developers spot performance issues more precisely and optimize their applications. The CPU-time profiling feature is experimental for now but could expand to other platforms later.

Another notable addition is support for PEM encodings of cryptographic objects. This API makes it easier to encode and decode cryptographic keys, certificates, and revocation lists in the widely used PEM format. Previously, working with PEM in Java was complicated, but now developers can handle PEM data more simply and reliably, supporting standard binary formats like PKCS#8 and X.509.

Enhancements for Performance and Memory

The Shenandoah garbage collector (GC) has moved from a preview feature to a fully supported product. This GC is designed to improve memory management, reduce pause times, and handle demanding workloads better. It features a generational mode, which optimizes collection efficiency by dividing objects based on their age. This update aims to make Shenandoah more stable and reliable for large, complex applications.

Another big focus is startup speed. Ahead-of-time (AOT) command-line ergonomics now make it easier to create AOT caches that speed up application launch. This means Java apps can start faster without changing existing workflows. Similarly, AOT method profiling allows the JVM to generate optimized code right from startup, improving warmup times without requiring code modifications. These features are meant to help Java apps become more responsive, especially in environments where quick startup is crucial.

Better Tools for Developers and AI Integration

Oracle has also introduced several features aimed at making Java more suitable for AI and data-heavy workloads. The vector API, commonly used in AI computations, is now more flexible and easier to use. Scoped values, which enable sharing immutable data efficiently across threads, are being previewed again. They offer a simpler alternative to thread-local variables, especially useful in modern Java applications that use virtual threads and structured concurrency.

Pattern matching with primitive types is also getting a significant upgrade. The third preview allows primitive types to be used in pattern matching, instanceof checks, and switch statements. This makes Java coding more consistent and expressive, especially when working with data from AI models. The goal is to make Java easier and more powerful for developers working on complex logic, including AI algorithms.

The compact object headers feature, which reduces the size of object metadata, is now a standard part of Java. Smaller headers mean less memory use and higher data locality, which can lead to better performance and denser deployment. Since its initial preview, this feature has proven stable and beneficial in real-world applications.

In summary, Java 25 brings a range of improvements that boost performance, security, and developer productivity. The new profiling and memory management tools are particularly helpful for AI and high-performance applications. As an LTS release, it’s set to support Java developers for years to come, making it a smart upgrade for many projects.

Inspired by

• https://www.infoworld.com/article/3846172/jdk-25-the-new-features-in-java-25.html

Sources

• gmpg.org
• pexels.com
• oracle.com
• java.net
• openjdk.org