Matrix Transpose on CUDA-LLM task suite benchmark leaderboard · SOTA2 Research

Matrix Transpose on CUDA-LLM task suite

5.2TimeOptiML

Evaluation Results

Method	Links
OptiML2026.02		5.2	1.79	15.77	76.31	23.12	718	1,587	379	61.92	26
HPC-Coder-V2 + OptiML-X2026.02		5.52	1.68	14.98	77.43	22.61	742	1,609	388	61.01	26
GPT 5.1 + OptiML-X2026.02		5.74	1.62	14.61	78.14	22.33	765	1,643	393	60.21	26
StarCoder2 + OptiML-X2026.02		5.83	1.6	14.13	78.96	21.84	776	1,679	397	59.84	26
Qwen2.5-Coder + OptiML-X2026.02		6	1.55	13.74	79.62	21.27	794	1,732	401	58.69	26
GPT 5.12026.02		9.3	—	10.22	86.47	18.31	973	2,108	417	52.33	24
Qwen2.5-Coder2026.02		9.6	0.97	9.88	87.22	17.95	1,004	2,196	434	51.76	24
HPC-Coder-V22026.02		9.88	0.94	9.61	87.58	17.72	1,019	2,234	441	51.33	24
StarCoder22026.02		10.21	0.91	9.32	87.96	17.48	1,045	2,288	453	50.89	26