DeepSeek V4 Pro 1.6T — B300 vs GB200 NVL72
Head-to-head AI inference benchmark comparison of B300 (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.
Near the low end of the 6–203 tok/s/user interactivity band, at 55 tok/s/user on DeepSeek V4 Pro 1.6T: B300 runs 4080 tok/s/GPU at $0.16/M tokens, GB200 NVL72 runs 9497 at $0.06/M. GB200 NVL72 is 153% cheaper per token; GB200 NVL72 delivers 133% more tok/s/GPU.
Setting 105 tok/s/user as the target on DeepSeek V4 Pro 1.6T, B300 produces 1609 tok/s/GPU ($0.40 per million tokens) and GB200 NVL72 produces 6644 ($0.09). GB200 NVL72 is 335% cheaper per token; GB200 NVL72 delivers 313% more tok/s/GPU.
At 154 tok/s/user interactivity on DeepSeek V4 Pro 1.6T, B300 delivers 908 tok/s/GPU at $0.72 per million tokens; GB200 NVL72 delivers 526 tok/s/GPU at $1.17. B300 is 63% cheaper per token; B300 delivers 73% more tok/s/GPU at this point. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:4079.5GB200 NVL72:9497.1 | B300:1608.8GB200 NVL72:6644.2 | B300:908.1GB200 NVL72:525.8 |
| Cost ($/M tok) | B300:$0.163GB200 NVL72:$0.065 | B300:$0.403GB200 NVL72:$0.093 | B300:$0.716GB200 NVL72:$1.169 |
| tok/s/MW | B300:1879965GB200 NVL72:4522446 | B300:741368GB200 NVL72:3163896 | B300:418488GB200 NVL72:250366 |
| Concurrency | B300:~153GB200 NVL72:~16349 | B300:~9GB200 NVL72:~2152 | B300:~3GB200 NVL72:~45 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.