Qwen 3.5 397B-A17B — GB300 NVL72 vs MI325X
Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) on Qwen 3.5 397B-A17B. 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 47 tok/s/user interactivity on Qwen 3.5 397B-A17B, GB300 NVL72 delivers 13149 tok/s/chip at $0.05 per million tokens; MI325X delivers 799 tok/s/chip at $0.38. GB300 NVL72 is 683% cheaper per token; GB300 NVL72 delivers 1546% more tok/s/chip at this point.
GB300 NVL72 posts 13149 tok/s/chip for $0.05 per million tokens at 54 tok/s/user on Qwen 3.5 397B-A17B; MI325X posts 601 tok/s/chip for $0.50. GB300 NVL72 is 922% cheaper per token; GB300 NVL72 delivers 2089% more tok/s/chip.
Throughput at 61 tok/s/user on Qwen 3.5 397B-A17B: GB300 NVL72 hits 12570 tok/s/chip, MI325X hits 433. Per-million costs land at $0.05 and $0.72 respectively. GB300 NVL72 is 1315% cheaper per token; GB300 NVL72 delivers 2803% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:13149.0MI325X:798.9 | GB300 NVL72:13149.0MI325X:600.6 | GB300 NVL72:12570.1MI325X:433.0 |
| Cost ($/M tok) | GB300 NVL72:$0.049MI325X:$0.382 | GB300 NVL72:$0.049MI325X:$0.499 | GB300 NVL72:$0.051MI325X:$0.722 |
| tok/s/MW | GB300 NVL72:6202353MI325X:472744 | GB300 NVL72:6202353MI325X:355407 | GB300 NVL72:5929270MI325X:256207 |
| Concurrency | GB300 NVL72:~2304MI325X:~16 | GB300 NVL72:~2304MI325X:~11 | GB300 NVL72:~1635MI325X:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.