Qwen 3.5 397B-A17B — H200 vs RTX PRO 6000
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and RTX PRO 6000 (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.
H200: 505 tok/s/GPU, $0.78 per million tokens at 65 tok/s/user on Qwen 3.5 397B-A17B. RTX PRO 6000 is unmeasured here.
At 105 tok/s/user on Qwen 3.5 397B-A17B, H200 delivers 353 tok/s/GPU at $1.10 per million tokens; RTX PRO 6000 hasn't been benchmarked at this target.
H200 hits 284 tok/s/GPU for $1.37 per million tokens at 144 tok/s/user on Qwen 3.5 397B-A17B. No RTX PRO 6000 data at this operating point. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:505.3RTX PRO 6000:— | H200:353.3RTX PRO 6000:— | H200:284.4RTX PRO 6000:— |
| Cost ($/M tok) | H200:$0.775RTX PRO 6000:— | H200:$1.096RTX PRO 6000:— | H200:$1.374RTX PRO 6000:— |
| tok/s/MW | H200:368816RTX PRO 6000:— | H200:257863RTX PRO 6000:— | H200:207582RTX PRO 6000:— |
| Concurrency | H200:~31RTX PRO 6000:— | H200:~14RTX PRO 6000:— | H200:~8RTX PRO 6000:— |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.