AIโ
โ
โ
โ
Hacker News ยท 2026-07-03
GLM-5.2 Achieves 2626 tok/s/node on AMD MI355X at Over 2x Lower CostWafer.ai has released a performance test of GLM-5.2 on AMD MI355X, showcasing impressive speed and cost advantages. This configuration achieves 2626 tok/s per node while reducing costs by over 2x compared to Blackwell. The results highlight the benefits of using AMD MI355X for GLM-5.2 workloads.
- GLM-5.2 achieves 2626 tok/s/node on AMD MI355X
- Cost is reduced by over 2x compared to Blackwell
- Wafer.ai's test results demonstrate the advantages of AMD MI355X