NVIDIA ComputeEval
NVIDIA benchmark for evaluating LLM-generated CUDA and CUDA Python code on correctness and optional GPU performance across kernels, runtime APIs, memory management, parallel algorithms, and GPU libraries.
0rows
pass_rateprimary metric
—sampled
Metadata
Metrics
Pass Rate, Speedup vs Baseline
| Rank | Subject | Pass Rate | Model Match | Provenance | Sampled |
|---|
No matching rows.