Qwen 3.5 397B-A17B — B300 vs MI325X Performance per Dollar
Cost per million tokens of B300 (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.
On Qwen 3.5 397B-A17B at 46 tok/s/user, the per-million math comes out to $0.15 for B300 and $0.86 for MI325X; B300 delivers 472% more output per dollar.
At 55 tok/s/user on Qwen 3.5 397B-A17B, B300 costs $0.18 per million tokens; MI325X costs $1.63. B300 is 811% more cost-efficient at this operating point.
B300 edges MI325X at 64 tok/s/user on Qwen 3.5 397B-A17B — $0.21 per million tokens versus $2.64, a 1170% cost-per-token gap. (Numbers reflect the default 1k/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): B300 $2.34/GPU/hr · MI325X $1.28/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 | B300:$0.151MI325X:$0.864 | B300:$0.179MI325X:$1.631 | B300:$0.208MI325X:$2.638 |
| Concurrency | B300:~205MI325X:~37 | B300:~140MI325X:~16 | B300:~99MI325X:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.