DeepSeek V4 Pro 1.6T — B200 vs GB200 NVL72
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) 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.
B200 posts 2976 tok/s/GPU for $0.18 per million tokens at 43 tok/s/user on DeepSeek V4 Pro 1.6T; GB200 NVL72 posts 2401 tok/s/GPU for $0.26. B200 is 42% cheaper per token; B200 delivers 24% more tok/s/GPU.
Throughput at 79 tok/s/user on DeepSeek V4 Pro 1.6T: B200 hits 822 tok/s/GPU, GB200 NVL72 hits 583. Per-million costs land at $0.67 and $1.05 respectively. B200 is 57% cheaper per token; B200 delivers 41% more tok/s/GPU.
B200 / GB200 NVL72 on DeepSeek V4 Pro 1.6T at 116 tok/s/user: 520 / 195 tok/s/GPU, $1.04 / $3.37 per million tokens. B200 is 225% cheaper per token; B200 delivers 166% more tok/s/GPU. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:2976.2GB200 NVL72:2401.3 | B200:822.3GB200 NVL72:582.9 | B200:519.6GB200 NVL72:195.4 |
| Cost ($/M tok) | B200:$0.180GB200 NVL72:$0.257 | B200:$0.668GB200 NVL72:$1.051 | B200:$1.036GB200 NVL72:$3.368 |
| tok/s/MW | B200:1371519GB200 NVL72:1143485 | B200:378939GB200 NVL72:277571 | B200:239428GB200 NVL72:93050 |
| Concurrency | B200:~74GB200 NVL72:~125 | B200:~10GB200 NVL72:~47 | B200:~4GB200 NVL72:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.