DeepSeek V4 Pro 1.6T — B200 vs B300
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and B300 (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.
At 44 tok/s/user interactivity on DeepSeek V4 Pro 1.6T, B200 delivers 1316 tok/s/GPU at $0.41 per million tokens; B300 delivers 837 tok/s/GPU at $0.77. B200 is 88% cheaper per token; B200 delivers 57% more tok/s/GPU at this point.
B200 posts 230 tok/s/GPU for $2.35 per million tokens at 83 tok/s/user on DeepSeek V4 Pro 1.6T; B300 posts 383 tok/s/GPU for $1.71. B300 is 37% cheaper per token; B300 delivers 67% more tok/s/GPU.
Throughput at 122 tok/s/user on DeepSeek V4 Pro 1.6T: B200 hits 100 tok/s/GPU, B300 hits 155. Per-million costs land at $5.43 and $4.28 respectively. B300 is 27% cheaper per token; B300 delivers 55% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:1316.1B300:837.4 | B200:229.8B300:383.3 | B200:99.6B300:154.6 |
| Cost ($/M tok) | B200:$0.413B300:$0.774 | B200:$2.348B300:$1.714 | B200:$5.429B300:$4.279 |
| tok/s/MW | B200:606506B300:385916 | B200:105900B300:176629 | B200:45910B300:71226 |
| Concurrency | B200:~137B300:~41 | B200:~12B300:~10 | B200:~4B300:~2 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.