Qwen 3.5 397B-A17B — H100 vs H200 Performance per Dollar
Cost per million tokens of H100 (NVIDIA Hopper) versus H200 (NVIDIA Hopper) 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 61 tok/s/user, the per-million math comes out to $1.43 for H100 and $0.93 for H200; H200 delivers 54% more output per dollar.
At 78 tok/s/user on Qwen 3.5 397B-A17B, H100 costs $1.78 per million tokens; H200 costs $1.19. H200 is 50% more cost-efficient at this operating point.
H200 edges H100 at 94 tok/s/user on Qwen 3.5 397B-A17B — $1.36 per million tokens versus $2.38, a 76% 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): H100 $1.30/GPU/hr · H200 $1.41/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 | H100:$1.429H200:$0.928 | H100:$1.779H200:$1.188 | H100:$2.381H200:$1.355 |
| Concurrency | H100:~17H200:~28 | H100:~10H200:~17 | H100:~7H200:~13 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.