Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) on DeepSeek V4.1 Flash 552B. 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.
AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →
Setting 92 tok/s/user as the target on DeepSeek V4.1 Flash 552B, GB300 NVL72 produces 121163 tok/s/chip ($0.01 per million tokens) and H200 produces 17798 ($0.02). GB300 NVL72 is 260% cheaper per token; GB300 NVL72 delivers 581% more tok/s/chip.
At 168 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, GB300 NVL72 delivers 45910 tok/s/chip at $0.01 per million tokens; H200 delivers 10229 tok/s/chip at $0.03. GB300 NVL72 is 137% cheaper per token; GB300 NVL72 delivers 349% more tok/s/chip at this point.
GB300 NVL72 posts 27605 tok/s/chip for $0.02 per million tokens at 244 tok/s/user on DeepSeek V4.1 Flash 552B; H200 posts 5418 tok/s/chip for $0.06. GB300 NVL72 is 169% cheaper per token; GB300 NVL72 delivers 409% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:121162.9H200:17798.0 | GB300 NVL72:45910.3H200:10229.2 | GB300 NVL72:27604.8H200:5418.2 |
| Cost ($/M tok) | GB300 NVL72:$0.005H200:$0.019 | GB300 NVL72:$0.014H200:$0.033 | GB300 NVL72:$0.023H200:$0.063 |
| tok/s/MW | GB300 NVL72:57152328H200:12991231 | GB300 NVL72:21655788H200:7466599 | GB300 NVL72:13021112H200:3954867 |
| Concurrency | GB300 NVL72:~47H200:~12 | GB300 NVL72:~14H200:~6 | GB300 NVL72:~7H200:~2 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
