Qwen 3.5 397B-A17B — GB300 NVL72 vs MI325X Performance per Dollar
Cost per million tokens of GB300 NVL72 (NVIDIA Blackwell) versus MI325X (AMD CDNA 3) 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.
GB300 NVL72 edges MI325X at 47 tok/s/user on Qwen 3.5 397B-A17B — $0.05 per million tokens versus $0.38, a 683% cost-per-token gap.
Push Qwen 3.5 397B-A17B to 54 tok/s/user and GB300 NVL72 lands at $0.05 per million tokens against MI325X's $0.50 — GB300 NVL72 pulls ahead by 922%.
GB300 NVL72: $0.05 per million tokens. MI325X: $0.72. Both at 61 tok/s/user on Qwen 3.5 397B-A17B, with GB300 NVL72 1315% 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.)
Chip pricing (owning hyperscaler): GB300 NVL72 $2.31/chip/hr · MI325X $1.10/chip/hr. Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | GB300 NVL72:$0.049MI325X:$0.382 | GB300 NVL72:$0.049MI325X:$0.499 | GB300 NVL72:$0.051MI325X:$0.722 |
| 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.