Head-to-head AI inference benchmark comparison of GB200 NVL72 (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.
At 65 tok/s/user interactivity on DeepSeek R1, GB200 NVL72 delivers 6057 tok/s/chip at $0.09 per million tokens; H200 delivers 1065 tok/s/chip at $0.32. GB200 NVL72 is 273% cheaper per token; GB200 NVL72 delivers 469% more tok/s/chip at this point.
GB200 NVL72 posts 4759 tok/s/chip for $0.11 per million tokens at 99 tok/s/user on DeepSeek R1; H200 posts 746 tok/s/chip for $0.45. GB200 NVL72 is 319% cheaper per token; GB200 NVL72 delivers 538% more tok/s/chip.
Throughput at 133 tok/s/user on DeepSeek R1: GB200 NVL72 hits 2941 tok/s/chip, H200 hits 588. Per-million costs land at $0.18 and $0.58 respectively. GB200 NVL72 is 228% cheaper per token; GB200 NVL72 delivers 400% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:6056.9H200:1064.8 | GB200 NVL72:4758.6H200:745.7 | GB200 NVL72:2941.1H200:587.6 |
| Cost ($/M tok) | GB200 NVL72:$0.085H200:$0.318 | GB200 NVL72:$0.109H200:$0.454 | GB200 NVL72:$0.176H200:$0.577 |
| tok/s/MW | GB200 NVL72:3238968H200:777216 | GB200 NVL72:2544727H200:544275 | GB200 NVL72:1572785H200:428936 |
| Concurrency | GB200 NVL72:~975H200:~155 | GB200 NVL72:~333H200:~7 | GB200 NVL72:~242H200:~4 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr:
Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
DeepSeek R1 0528 671B 8K / 1K Total Tokens per $1 TCO vs. Interactivity