From the docs
Improved External Transforms Performance with Lightweight Mode | External transforms now supports the @lightweight API, resulting in improved performance by reducing startup overhead. This enhancement is ideal for external transforms as they are typically not Spark dependent or CPU-intensive, leading to better latency and throughput.
Why this matters
For most datasets in most enterprises, you probably don’t need the full power of Spark. Lightweight Transforms typically have less overhead and can offer performance benefits on small data.
The ability to hit external endpoints from within lightweight transforms means that they’re one step closer to being a full replacement for Spark-based transforms for most use cases.