Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and H100 (NVIDIA Hopper) on gpt-oss 120B. 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 113 tok/s/user interactivity on gpt-oss 120B, GB200 NVL72 delivers 42816 tok/s/chip at $0.01 per million tokens; H100 delivers 5280 tok/s/chip at $0.06. GB200 NVL72 is 410% cheaper per token; GB200 NVL72 delivers 711% more tok/s/chip at this point.
GB200 NVL72 posts 32911 tok/s/chip for $0.02 per million tokens at 159 tok/s/user on gpt-oss 120B; H100 posts 3672 tok/s/chip for $0.09. GB200 NVL72 is 464% cheaper per token; GB200 NVL72 delivers 796% more tok/s/chip.
Throughput at 205 tok/s/user on gpt-oss 120B: GB200 NVL72 hits 25756 tok/s/chip, H100 hits 2428. Per-million costs land at $0.02 and $0.13 respectively. GB200 NVL72 is 567% cheaper per token; GB200 NVL72 delivers 961% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:42815.7H100:5279.9 | GB200 NVL72:32910.9H100:3671.7 | GB200 NVL72:25755.8H100:2428.1 |
| Cost ($/M tok) | GB200 NVL72:$0.012H100:$0.062 | GB200 NVL72:$0.016H100:$0.089 | GB200 NVL72:$0.020H100:$0.134 |
| tok/s/MW | GB200 NVL72:22896080H100:3853936 | GB200 NVL72:17599435H100:2680055 | GB200 NVL72:13773170H100:1772304 |
| Concurrency | GB200 NVL72:~375H100:~25 | GB200 NVL72:~291H100:~11 | GB200 NVL72:~87H100:~5 |
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
gpt-oss 120B 8K / 1K Total Tokens per $1 TCO vs. Interactivity