Qwen 3.5 397B-A17B — B200 vs GB300 NVL72
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB300 NVL72 (NVIDIA Blackwell) 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.
Throughput at 90 tok/s/user on Qwen 3.5 397B-A17B: B200 hits 9634 tok/s/GPU, GB300 NVL72 hits 6025. Per-million costs land at $0.06 and $0.12 respectively. B200 is 117% cheaper per token; B200 delivers 60% more tok/s/GPU.
B200 / GB300 NVL72 on Qwen 3.5 397B-A17B at 136 tok/s/user: 6507 / 2801 tok/s/GPU, $0.08 / $0.26 per million tokens. B200 is 211% cheaper per token; B200 delivers 132% more tok/s/GPU.
Toward the upper edge of the 44–228 tok/s/user interactivity band, at 183 tok/s/user on Qwen 3.5 397B-A17B: B200 runs 4561 tok/s/GPU at $0.12/M tokens, GB300 NVL72 runs 1338 at $0.55/M. B200 is 359% cheaper per token; B200 delivers 241% more tok/s/GPU. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:9633.6GB300 NVL72:6024.7 | B200:6506.5GB300 NVL72:2801.0 | B200:4560.7GB300 NVL72:1338.4 |
| Cost ($/M tok) | B200:$0.056GB300 NVL72:$0.122 | B200:$0.082GB300 NVL72:$0.256 | B200:$0.121GB300 NVL72:$0.553 |
| tok/s/MW | B200:4439432GB300 NVL72:2868920 | B200:2998402GB300 NVL72:1333831 | B200:2101694GB300 NVL72:637343 |
| Concurrency | B200:~114GB300 NVL72:~71 | B200:~12GB300 NVL72:~21 | B200:~6GB300 NVL72:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.