How do you optimize Quarkus native image build times?

Native-image builds are inherently slower than a normal compile because of the static analysis GraalVM performs, but several concrete levers reduce build time without giving up the benefits of native compilation in production.

• Use Dev Mode / JVM mode for daily development: only build natively for final validation and CI/CD, not on every local code change.
• Allocate sufficient build memory: an under-resourced build machine causes swapping or failure, which is often slower (or fatal) compared to giving the build adequate RAM upfront.
• Use quick builds during iteration: the quarkus.native.additional-build-args=-Ob (quick build mode) trades some runtime performance for significantly faster native builds, useful when iterating on native-specific issues.
• Build inside a container matching the target OS only when needed for cross-platform correctness, since container-based builds add overhead compared to a native build on a matching host OS directly.
• Minimize unnecessary dependencies: fewer classes for GraalVM's static analysis to reason about generally means a faster build, so trimming unused libraries has a compounding effect.

In CI/CD specifically, caching GraalVM's own installation and reusing build agents with warm OS-level caches (rather than a fully clean container every run) also meaningfully cuts wall-clock time across repeated builds, since a large part of the fixed cost is GraalVM setup rather than the application-specific analysis itself.