Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and H100 (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 →
Near the low end of the 16–303 tok/s/user interactivity band, at 88 tok/s/user on DeepSeek V4.1 Flash 552B: GB300 NVL72 runs 125508 tok/s/chip at $0.01/M tokens, H100 runs 10466 at $0.03/M. GB300 NVL72 is 507% cheaper per token; GB300 NVL72 delivers 1099% more tok/s/chip.
Setting 160 tok/s/user as the target on DeepSeek V4.1 Flash 552B, GB300 NVL72 produces 49336 tok/s/chip ($0.01 per million tokens) and H100 produces 6697 ($0.05). GB300 NVL72 is 273% cheaper per token; GB300 NVL72 delivers 637% more tok/s/chip.
At 232 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, GB300 NVL72 delivers 28701 tok/s/chip at $0.02 per million tokens; H100 delivers 4131 tok/s/chip at $0.08. GB300 NVL72 is 252% cheaper per token; GB300 NVL72 delivers 595% more tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:125507.6H100:10466.0 | GB300 NVL72:49336.1H100:6696.7 | GB300 NVL72:28701.0H100:4131.2 |
| Cost ($/M tok) | GB300 NVL72:$0.005H100:$0.031 | GB300 NVL72:$0.013H100:$0.049 | GB300 NVL72:$0.022H100:$0.079 |
| tok/s/MW | GB300 NVL72:59201687H100:7639404 | GB300 NVL72:23271748H100:4888077 | GB300 NVL72:13538194H100:3015510 |
| Concurrency | GB300 NVL72:~49H100:~14 | GB300 NVL72:~15H100:~7 | GB300 NVL72:~8H100:~4 |
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.
