DeepSeek V4 Pro 1.6T — GB300 NVL72 vs H200
Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) on DeepSeek V4 Pro 1.6T. 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 hits 211 tok/s/GPU for $1.85 per million tokens at 45 tok/s/user on DeepSeek V4 Pro 1.6T. No GB300 NVL72 data at this operating point.
H200: 92 tok/s/GPU, $4.26 per million tokens at 87 tok/s/user on DeepSeek V4 Pro 1.6T. GB300 NVL72 is unmeasured here.
At 128 tok/s/user on DeepSeek V4 Pro 1.6T, H200 delivers 61 tok/s/GPU at $6.36 per million tokens; GB300 NVL72 hasn't been benchmarked at this target. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB300 NVL72:—H200:211.4 | GB300 NVL72:—H200:92.3 | GB300 NVL72:—H200:61.2 |
| Cost ($/M tok) | GB300 NVL72:—H200:$1.851 | GB300 NVL72:—H200:$4.262 | GB300 NVL72:—H200:$6.362 |
| tok/s/MW | GB300 NVL72:—H200:122223 | GB300 NVL72:—H200:53359 | GB300 NVL72:—H200:35354 |
| Concurrency | GB300 NVL72:—H200:~20 | GB300 NVL72:—H200:~5 | GB300 NVL72:—H200:~2 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.