Qwen 3.5 397B-A17B — GB300 NVL72 vs RTX PRO 6000
Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and RTX PRO 6000 (NVIDIA Blackwell) 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.
GB300 NVL72 / RTX PRO 6000 on Qwen 3.5 397B-A17B at 61 tok/s/user: 23235 / 931 tok/s/chip, $0.03 / $0.19 per million tokens. GB300 NVL72 is 601% cheaper per token; GB300 NVL72 delivers 2395% more tok/s/chip.
Around the middle of the 31–153 tok/s/user interactivity band, at 92 tok/s/user on Qwen 3.5 397B-A17B: GB300 NVL72 runs 21121 tok/s/chip at $0.03/M tokens, RTX PRO 6000 runs 726 at $0.26/M. GB300 NVL72 is 756% cheaper per token; GB300 NVL72 delivers 2808% more tok/s/chip.
Setting 123 tok/s/user as the target on Qwen 3.5 397B-A17B, GB300 NVL72 produces 18398 tok/s/chip ($0.04 per million tokens) and RTX PRO 6000 produces 505 ($0.41). GB300 NVL72 is 1072% cheaper per token; GB300 NVL72 delivers 3546% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:23235.2RTX PRO 6000:931.3 | GB300 NVL72:21120.7RTX PRO 6000:726.3 | GB300 NVL72:18397.7RTX PRO 6000:504.6 |
| Cost ($/M tok) | GB300 NVL72:$0.027RTX PRO 6000:$0.191 | GB300 NVL72:$0.030RTX PRO 6000:$0.260 | GB300 NVL72:$0.035RTX PRO 6000:$0.412 |
| tok/s/MW | GB300 NVL72:10960012RTX PRO 6000:955182 | GB300 NVL72:9962582RTX PRO 6000:744882 | GB300 NVL72:8678181RTX PRO 6000:517546 |
| Concurrency | GB300 NVL72:~1536RTX PRO 6000:~8 | GB300 NVL72:~1535RTX PRO 6000:~4 | GB300 NVL72:~1068RTX PRO 6000:~2 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.