DeepSeek R1 — B300 vs GB200 NVL72
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) on DeepSeek R1. 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.
B300 / GB200 NVL72 on DeepSeek R1 at 93 tok/s/user: 8128 / 11532 tok/s/chip, $0.08 / $0.04 per million tokens. GB200 NVL72 is 72% cheaper per token; GB200 NVL72 delivers 42% more tok/s/chip.
Around the middle of the 24–299 tok/s/user interactivity band, at 162 tok/s/user on DeepSeek R1: B300 runs 1801 tok/s/chip at $0.35/M tokens, GB200 NVL72 runs 4192 at $0.12/M. GB200 NVL72 is 182% cheaper per token; GB200 NVL72 delivers 133% more tok/s/chip.
Setting 231 tok/s/user as the target on DeepSeek R1, B300 produces 1055 tok/s/chip ($0.59 per million tokens) and GB200 NVL72 produces 1012 ($0.51). GB200 NVL72 is 15% cheaper per token; B300 delivers 4% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:8127.6GB200 NVL72:11531.8 | B300:1800.8GB200 NVL72:4192.5 | B300:1054.8GB200 NVL72:1012.2 |
| Cost ($/M tok) | B300:$0.077GB200 NVL72:$0.045 | B300:$0.348GB200 NVL72:$0.123 | B300:$0.589GB200 NVL72:$0.514 |
| tok/s/MW | B300:4277680GB200 NVL72:6166725 | B300:947790GB200 NVL72:2241975 | B300:555139GB200 NVL72:541259 |
| Concurrency | B300:~194GB200 NVL72:~913 | B300:~45GB200 NVL72:~178 | B300:~33GB200 NVL72:~19 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.