gpt-oss 120B — GB200 NVL72 vs MI325X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) 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.
GB200 NVL72 posts 50567 tok/s/GPU for $0.01 per million tokens at 51 tok/s/user on gpt-oss 120B; MI325X posts 2704 tok/s/GPU for $0.13. GB200 NVL72 is 979% cheaper per token; GB200 NVL72 delivers 1770% more tok/s/GPU.
Throughput at 72 tok/s/user on gpt-oss 120B: GB200 NVL72 hits 45654 tok/s/GPU, MI325X hits 1889. Per-million costs land at $0.01 and $0.19 respectively. GB200 NVL72 is 1316% cheaper per token; GB200 NVL72 delivers 2317% more tok/s/GPU.
GB200 NVL72 / MI325X on gpt-oss 120B at 92 tok/s/user: 36227 / 983 tok/s/GPU, $0.02 / $0.35 per million tokens. GB200 NVL72 is 1970% cheaper per token; GB200 NVL72 delivers 3586% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:50567.3MI325X:2704.4 | GB200 NVL72:45653.7MI325X:1888.9 | GB200 NVL72:36227.0MI325X:982.7 |
| Cost ($/M tok) | GB200 NVL72:$0.012MI325X:$0.131 | GB200 NVL72:$0.013MI325X:$0.190 | GB200 NVL72:$0.017MI325X:$0.351 |
| tok/s/MW | GB200 NVL72:24079678MI325X:1240543 | GB200 NVL72:21739878MI325X:866471 | GB200 NVL72:17250936MI325X:450777 |
| Concurrency | GB200 NVL72:~1559MI325X:~64 | GB200 NVL72:~1026MI325X:~60 | GB200 NVL72:~1345MI325X:~22 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.