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 50 tok/s/user interactivity on DeepSeek V4 Pro 1.6T, B200 delivers 2707 tok/s/GPU at $0.20 per million tokens; B300 delivers 1952 tok/s/GPU at $0.34. B200 is 71% cheaper per token; B200 delivers 39% more tok/s/GPU at this point.
B200 posts 237 tok/s/GPU for $2.34 per million tokens at 94 tok/s/user on DeepSeek V4 Pro 1.6T; B300 posts 391 tok/s/GPU for $1.63. B300 is 43% cheaper per token; B300 delivers 65% more tok/s/GPU.
Throughput at 138 tok/s/user on DeepSeek V4 Pro 1.6T: B200 hits 92 tok/s/GPU, B300 hits 159. Per-million costs land at $5.69 and $4.14 respectively. B300 is 37% cheaper per token; B300 delivers 73% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:2707.4B300:1951.7 | B200:237.0B300:390.7 | B200:92.2B300:159.4 |
| Cost ($/M tok) | B200:$0.200B300:$0.343 | B200:$2.337B300:$1.633 | B200:$5.685B300:$4.142 |
| tok/s/MW | B200:1247661B300:899383 | B200:109213B300:180058 | B200:42486B300:73474 |
| Concurrency | B200:~261B300:~95 | B200:~12B300:~9 | B200:~3B300:~3 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.