We will discuss the strengths and tradeoffs of a subset of GPU-focused crates and frameworks along with insights from evaluating scientific computing benchmarks.