Breaking: Go 1.25 Introduces Green Tea Garbage Collector
Go 1.25 includes a new experimental garbage collector named Green Tea, designed to dramatically reduce the time applications spend on memory management. Early benchmarks show that many workloads experience a 10% reduction in garbage collection (GC) CPU time, while some see improvements of up to 40%.
The feature is activated by setting GOEXPERIMENT=greenteagc at build time. Google has already deployed it in production environments, signaling that the runtime is stable enough for real-world testing. The Go team urges developers to try it and report results, as feedback will guide the next steps.
Quotes from the Go Team
“Green Tea represents a major leap forward in Go runtime efficiency, especially for memory-intensive workloads,” said Michael Knyszek, Go team member at Google. “We’ve seen workloads where GC time dropped by nearly half, and we believe this will become the standard for Go applications.”
Austin Clements, also on the Go team, added: “We’re excited to share this with the community. Your real-world testing is critical—please file issues for problems and successes alike. We plan to make Green Tea the default collector in Go 1.26.”
Background: How Garbage Collection Works in Go
Garbage collection (GC) automatically reclaims memory no longer used by a program, freeing developers from manual memory management. In Go, the GC focuses on objects—values allocated on the heap—and pointers that reference them.
When the Go compiler cannot determine another way to allocate a value’s memory, the value becomes a heap object. The collector traces these objects to identify which are still reachable and then recycles the rest. Green Tea improves this tracing process, leading to lower CPU overhead and shorter pause times.
The new algorithm builds on Go’s existing concurrent, tri-color mark-and-sweep collector but introduces optimizations that reduce the work needed during each cycle. While some workloads see negligible gains, others—particularly those with large heaps or many small allocations—benefit significantly.
What This Means for Go Developers
For most Go developers, Green Tea promises smoother application performance with less GC-related jitter. Even a 10% reduction in GC time can translate to higher throughput and lower latency, especially in services handling many requests.
Developers should test their applications with GOEXPERIMENT=greenteagc as soon as possible. If your workload shows improvement, you can expect even better performance when Green Tea becomes the default. For those who see no benefit or regressions, reporting issues will help the team refine the collector before the 1.26 release.
Google’s internal usage confirms production readiness. The team emphasizes that Green Tea is not a prototype—it’s already running critical services. However, because garbage collection behavior can vary widely, community testing is essential.
Call to Action
To report problems, file a new issue on the Go issue tracker. To share successes, reply to the existing Green Tea issue. Your feedback will directly influence the path to default.
For more technical details, see Michael Knyszek’s GopherCon 2025 talk, which provides an in-depth explanation of the Green Tea algorithm and its performance characteristics.
This is a developing story. The Go team expects to finalize the design by mid-2026 based on community feedback.