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 2711 tok/s/GPU for $0.13. GB200 NVL72 is 980% cheaper per token; GB200 NVL72 delivers 1765% more tok/s/GPU.
Throughput at 72 tok/s/user on gpt-oss 120B: GB200 NVL72 hits 45654 tok/s/GPU, MI325X hits 1999. Per-million costs land at $0.01 and $0.18 respectively. GB200 NVL72 is 1230% cheaper per token; GB200 NVL72 delivers 2184% more tok/s/GPU.
GB200 NVL72 / MI325X on gpt-oss 120B at 93 tok/s/user: 35620 / 1049 tok/s/GPU, $0.02 / $0.34 per million tokens. GB200 NVL72 is 1874% cheaper per token; GB200 NVL72 delivers 3297% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:50567.3MI325X:2711.1 | GB200 NVL72:45653.7MI325X:1998.6 | GB200 NVL72:35620.2MI325X:1048.7 |
| Cost ($/M tok) | GB200 NVL72:$0.012MI325X:$0.131 | GB200 NVL72:$0.013MI325X:$0.179 | GB200 NVL72:$0.017MI325X:$0.340 |
| tok/s/MW | GB200 NVL72:24079678MI325X:1243645 | GB200 NVL72:21739878MI325X:916809 | GB200 NVL72:16961997MI325X:481036 |
| Concurrency | GB200 NVL72:~1559MI325X:~64 | GB200 NVL72:~1026MI325X:~60 | GB200 NVL72:~1225MI325X:~25 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.