gpt-oss 120B · Chip comparison
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 52,037 tok/s/chip for $0.010 per million tokens at 78 tok/s/user on gpt-oss 120B; MI325X posts 3,123 tok/s/chip for $0.098. GB200 NVL72 is 90% cheaper per token; GB200 NVL72 delivers 17× the tok/s/chip.
Throughput at 87 tok/s/user on gpt-oss 120B: GB200 NVL72 hits 49,784 tok/s/chip, MI325X hits 1,675. Costs per million tokens land at $0.010 and $0.18 respectively. GB200 NVL72 is 94% cheaper per token; GB200 NVL72 delivers 30× the tok/s/chip.
GB200 NVL72 / MI325X on gpt-oss 120B at 97 tok/s/user: 47,222 / 1,070 tok/s/chip, $0.011 / $0.29 per million tokens. GB200 NVL72 is 96% cheaper per token; GB200 NVL72 delivers 44× the tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL7252,037MI325X3,123 | GB200 NVL7249,784MI325X1,675 | GB200 NVL7247,222MI325X1,070 |
| Cost ($/M tok) | GB200 NVL72$0.010MI325X$0.098 | GB200 NVL72$0.010MI325X$0.18 | GB200 NVL72$0.011MI325X$0.29 |
| tok/s/MW | GB200 NVL7227,827,087MI325X1,847,849 | GB200 NVL7226,622,665MI325X991,335 | GB200 NVL7225,252,347MI325X633,210 |
| Concurrency | GB200 NVL72~323MI325X~8 | GB200 NVL72~268MI325X~15 | GB200 NVL72~261MI325X~7 |