Qwen 3.5 397B-A17B — GB200 NVL72 vs GB300 NVL72 Performance per Dollar
Cost per million tokens of GB200 NVL72 (NVIDIA Blackwell) versus GB300 NVL72 (NVIDIA Blackwell) on Qwen 3.5 397B-A17B. Owning-hyperscaler TCO normalized by output tokens — performance per dollar across LLM workloads. Pick the more cost-efficient SKU at every target interactivity level. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
GB200 NVL72 edges GB300 NVL72 at 121 tok/s/user on Qwen 3.5 397B-A17B — $0.11 per million tokens versus $0.12, a 3% cost-per-token gap.
Push Qwen 3.5 397B-A17B to 203 tok/s/user and GB200 NVL72 lands at $0.33 per million tokens against GB300 NVL72's $0.36 — GB200 NVL72 pulls ahead by 7%.
GB200 NVL72: $0.71 per million tokens. GB300 NVL72: $0.86. Both at 284 tok/s/user on Qwen 3.5 397B-A17B, with GB200 NVL72 21% cheaper. (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.)
GPU pricing (owning hyperscaler): GB200 NVL72 $2.21/GPU/hr · GB300 NVL72 $2.65/GPU/hr. Source: SemiAnalysis Market August 2025 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | GB200 NVL72:$0.113GB300 NVL72:$0.117 | GB200 NVL72:$0.334GB300 NVL72:$0.358 | GB200 NVL72:$0.708GB300 NVL72:$0.860 |
| Concurrency | GB200 NVL72:~191GB300 NVL72:~241 | GB200 NVL72:~17GB300 NVL72:~23 | GB200 NVL72:~4GB300 NVL72:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.