Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and GB300 NVL72 (NVIDIA Blackwell) on DeepSeek V4.1 Flash 552B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics β 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 β
Setting 177 tok/s/user as the target on DeepSeek V4.1 Flash 552B, GB200 NVL72 produces 76605 tok/s/chip ($0.01 per million tokens) and GB300 NVL72 produces 686740 ($0.00). GB300 NVL72 is 622% cheaper per token; GB300 NVL72 delivers 796% more tok/s/chip.
At 305 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, GB200 NVL72 delivers 36774 tok/s/chip at $0.01 per million tokens; GB300 NVL72 delivers 29614 tok/s/chip at $0.02. GB200 NVL72 is 54% cheaper per token; GB200 NVL72 delivers 24% more tok/s/chip at this point.
GB200 NVL72 posts 15988 tok/s/chip for $0.03 per million tokens at 433 tok/s/user on DeepSeek V4.1 Flash 552B; GB300 NVL72 posts 27905 tok/s/chip for $0.02. GB300 NVL72 is 41% cheaper per token; GB300 NVL72 delivers 75% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:76604.7GB300 NVL72:686739.8 | GB200 NVL72:36774.4GB300 NVL72:29614.2 | GB200 NVL72:15988.5GB300 NVL72:27904.6 |
| Cost ($/M tok) | GB200 NVL72:$0.007GB300 NVL72:$0.001 | GB200 NVL72:$0.014GB300 NVL72:$0.022 | GB200 NVL72:$0.032GB300 NVL72:$0.023 |
| tok/s/MW | GB200 NVL72:40965059GB300 NVL72:323933881 | GB200 NVL72:19665440GB300 NVL72:13968983 | GB200 NVL72:8549988GB300 NVL72:13162539 |
| Concurrency | GB200 NVL72:~51GB300 NVL72:~3590 | GB200 NVL72:~20GB300 NVL72:~13 | GB200 NVL72:~8GB300 NVL72:~15 |
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.
