gpt-oss 120B — MI325X vs MI355X
Head-to-head AI inference benchmark comparison of MI325X (AMD CDNA 3) and MI355X (AMD CDNA 4) on gpt-oss 120B. 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 81 tok/s/user interactivity on gpt-oss 120B, MI325X delivers 1260 tok/s/GPU at $0.28 per million tokens; MI355X delivers 15234 tok/s/GPU at $0.03. MI355X is 935% cheaper per token; MI355X delivers 1109% more tok/s/GPU at this point.
MI325X posts 1076 tok/s/GPU for $0.33 per million tokens at 92 tok/s/user on gpt-oss 120B; MI355X posts 12995 tok/s/GPU for $0.03. MI355X is 945% cheaper per token; MI355X delivers 1108% more tok/s/GPU.
Throughput at 103 tok/s/user on gpt-oss 120B: MI325X hits 743 tok/s/GPU, MI355X hits 11108. Per-million costs land at $0.48 and $0.04 respectively. MI355X is 1201% cheaper per token; MI355X delivers 1395% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MI325X:1259.9MI355X:15234.5 | MI325X:1075.9MI355X:12995.3 | MI325X:743.0MI355X:11108.4 |
| Cost ($/M tok) | MI325X:$0.282MI355X:$0.027 | MI325X:$0.331MI355X:$0.032 | MI325X:$0.480MI355X:$0.037 |
| tok/s/MW | MI325X:577938MI355X:5748865 | MI325X:493524MI355X:4903883 | MI325X:340833MI355X:4191834 |
| Concurrency | MI325X:~19MI355X:~100 | MI325X:~26MI355X:~74 | MI325X:~15MI355X:~55 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.