Scikit-survival version 0.28.0 has been released, introducing support for Polars DataFrames alongside pandas DataFrames. This is achieved through the Narwhals dataframe abstraction layer, enhancing performance capabilities. The release also ensures compatibility with scikit-learn version 1.9. Datasets within scikit-survival can now be loaded as Polars DataFrames by specifying the output_type argument. Scikit-learn's set_output API allows transformers to return Polars DataFrames, and scikit-survival's ClinicalKernelTransform and OneHotEncoder now accept both Polars and pandas inputs. ClinicalKernelTransform is aware of Polars column types and allows specifying category order for ordinality. OneHotEncoder converts string columns to numeric ones, returning a DataFrame of the same type as the input. Survival models can now be fitted and predicted using Polars DataFrames, but only for eager DataFrames, not lazy ones. This release acknowledges two new contributors for their work on scikit-learn and Polars support. Updated dependencies include Narwhals 2.0.1 and scikit-learn 1.9.0. Scikit-survival can be installed via pip or conda.