GLM 5/5.1 — MI325X vs MI355X
Head-to-head AI inference benchmark comparison of MI325X (AMD CDNA 3) and MI355X (AMD CDNA 4) on GLM 5/5.1. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
At 21 tok/s/user interactivity on GLM 5/5.1, MI325X delivers 173 tok/s/GPU at $2.08 per million tokens; MI355X delivers 1025 tok/s/GPU at $0.41. MI355X is 414% cheaper per token; MI355X delivers 493% more tok/s/GPU at this point.
MI325X posts 101 tok/s/GPU for $3.53 per million tokens at 26 tok/s/user on GLM 5/5.1; MI355X posts 980 tok/s/GPU for $0.42. MI355X is 734% cheaper per token; MI355X delivers 868% more tok/s/GPU.
Throughput at 30 tok/s/user on GLM 5/5.1: MI325X hits 62 tok/s/GPU, MI355X hits 955. Per-million costs land at $5.71 and $0.43 respectively. MI355X is 1213% cheaper per token; MI355X delivers 1451% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MI325X:172.9MI355X:1025.3 | MI325X:101.3MI355X:980.4 | MI325X:61.6MI355X:954.6 |
| Cost ($/M tok) | MI325X:$2.081MI355X:$0.405 | MI325X:$3.534MI355X:$0.424 | MI325X:$5.710MI355X:$0.435 |
| tok/s/MW | MI325X:79293MI355X:386901 | MI325X:46466MI355X:369974 | MI325X:28234MI355X:360216 |
| Concurrency | MI325X:~34MI355X:~128 | MI325X:~16MI355X:~128 | MI325X:~9MI355X:~128 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.