This document benchmarks different implementations of linear regression in Pharo and compares them to implementations in Python. It finds that: 1) Calling the LAPACK library from Pharo provides a significant speedup of up to 2,103 times compared to a pure Pharo implementation. 2) The scikit-learn Python implementation is faster than the Pharo+LAPACK prototype by 8-5 times depending on dataset size. 3) A pure Pharo implementation is up to 15 times faster than an equivalent pure Python implementation. The results suggest Pharo can achieve high performance by leveraging optimized numerical libraries like LAPACK and lay the foundation for future work building fast machine learning libraries in Pharo.