It depends what you're measuring. For streaming workloads they're reporting 923x speedup for aggregations, and 105x for insertions, over iceberg [1].
There are also some improvements other areas, e.g. cross table transactions are not possible in iceberg/delta, but they are in ducklake. [2]
[1] https://ducklake.select/2026/04/02/data-inlining-in-ducklake...