DeepSeek R1 — B300 vs H200
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) 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.
Setting 50 tok/s/user as the target on DeepSeek R1, B300 produces 5852 tok/s/chip ($0.11 per million tokens) and H200 produces 1591 ($0.21). B300 is 96% cheaper per token; B300 delivers 268% more tok/s/chip.
At 89 tok/s/user interactivity on DeepSeek R1, B300 delivers 2298 tok/s/chip at $0.26 per million tokens; H200 delivers 819 tok/s/chip at $0.41. B300 is 59% cheaper per token; B300 delivers 181% more tok/s/chip at this point.
B300 posts 992 tok/s/chip for $0.64 per million tokens at 128 tok/s/user on DeepSeek R1; H200 posts 625 tok/s/chip for $0.55. H200 is 17% cheaper per token; B300 delivers 59% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:5852.2H200:1590.9 | B300:2298.5H200:819.4 | B300:992.2H200:624.5 |
| Cost ($/M tok) | B300:$0.108H200:$0.212 | B300:$0.261H200:$0.415 | B300:$0.642H200:$0.548 |
| tok/s/MW | B300:3080114H200:1161268 | B300:1209733H200:598071 | B300:522232H200:455850 |
| Concurrency | B300:~288H200:~73 | B300:~57H200:~14 | B300:~7H200:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.