PTXBench
Explore score history ↗PTXBench evaluates LLMs in generating architecture-specific PTX for GPU kernel optimization. It measures functional correctness, execution of target instructions, and speedup over frontier libraries across GEMM and attention workloads on H100 and B200 GPUs.
- Released
- 2026-08-18
- Institutions
- Stanford University · Carnegie Mellon University
- Evaluates
- Architecture-specific programming, Code, Code & Software, Code generation, GPU kernel optimization, Iterative code repair, Software & AI Compute
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- source-reviewed
PTXBench paper, code and dataset
No reported scores are on record for this benchmark yet.