DeepSeek R1 · Chip comparison
DeepSeek R1 — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (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.
Throughput at 51 tok/s/user on DeepSeek R1: B200 hits 4,764 tok/s/chip, H200 hits 1,499. Costs per million tokens land at $0.10 and $0.23 respectively. B200 is 55% cheaper per token; B200 delivers 3.2× the tok/s/chip.
B200 / H200 on DeepSeek R1 at 90 tok/s/user: 1,456 / 785 tok/s/chip, $0.33 / $0.43 per million tokens. B200 is 24% cheaper per token; B200 delivers 85% more tok/s/chip.
Toward the upper edge of the 13–167 tok/s/user interactivity band, at 129 tok/s/user on DeepSeek R1: B200 runs 1,044 tok/s/chip at $0.46 per million tokens, H200 runs 523 at $0.65. B200 is 29% cheaper per token; B200 delivers 100% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B2004,764H2001,499 | B2001,456H200785 | B2001,044H200523 |
| Cost ($/M tok) | B200$0.10H200$0.23 | B200$0.33H200$0.43 | B200$0.46H200$0.65 |
| tok/s/MW | B2002,785,967H2001,094,512 | B200851,700H200573,273 | B200610,699H200381,794 |
| Concurrency | B200~288H200~54 | B200~43H200~8 | B200~9H200~4 |