Qwen 3.5 397B-A17B — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) 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.
B200 posts 3811 tok/s/GPU for $0.14 per million tokens at 55 tok/s/user on Qwen 3.5 397B-A17B; H200 posts 463 tok/s/GPU for $0.85. B200 is 495% cheaper per token; B200 delivers 723% more tok/s/GPU.
Throughput at 81 tok/s/user on Qwen 3.5 397B-A17B: B200 hits 2426 tok/s/GPU, H200 hits 319. Per-million costs land at $0.22 and $1.23 respectively. B200 is 450% cheaper per token; B200 delivers 661% more tok/s/GPU.
B200 / H200 on Qwen 3.5 397B-A17B at 107 tok/s/user: 1564 / 271 tok/s/GPU, $0.34 / $1.45 per million tokens. B200 is 327% cheaper per token; B200 delivers 476% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:3810.6H200:462.9 | B200:2425.9H200:318.8 | B200:1564.4H200:271.4 |
| Cost ($/M tok) | B200:$0.142H200:$0.846 | B200:$0.223H200:$1.229 | B200:$0.338H200:$1.446 |
| tok/s/MW | B200:1756045H200:267580 | B200:1117918H200:184266 | B200:720928H200:156874 |
| Concurrency | B200:~155H200:~34 | B200:~65H200:~16 | B200:~33H200:~12 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.