For Delta Lake tables, a target file size of approximately 1 GB is recommended. This size provides an optimal balance between enabling efficient Spark task parallelism and minimizing metadata overhead, which can occur when managing a very large number of small files.