Benchmarking LLMs on CUDA with ComputeEval 2025.2

Benchmarking LLMs on CUDA with ComputeEval 2025.2