Prefix Sum on CUDA-LLM task suite
6.1TimeOptiML
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| OptiMLGPU=A100 80GB, Variant=G+X2026.02 | 6.1 | 1.09 | 29.96 | 40.22 | 27.33 | 813 | 1,842 | 1,371 | 74.15 | 42 | |
| HPC-Coder-V2 + OptiML-XGPU=A100 80GB2026.02 | 6.28 | 1.06 | 29.05 | 40.91 | 26.71 | 836 | 1,894 | 1,411 | 73.12 | 42 | |
| GPT 5.1 + OptiML-XGPU=A100 80GB2026.02 | 6.65 | — | 28.21 | 42.16 | 26.14 | 864 | 1,958 | 1,459 | 72.33 | 42 | |
| StarCoder2 + OptiML-XGPU=A100 80GB2026.02 | 6.74 | 0.99 | 27.55 | 42.66 | 25.53 | 878 | 1,989 | 1,478 | 71.14 | 42 | |
| Qwen2.5-Coder + OptiML-XGPU=A100 80GB2026.02 | 7.05 | 0.94 | 26.73 | 43.74 | 24.98 | 902 | 2,046 | 1,516 | 70.26 | 42 |