Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and Vera Rubin NVL72 (NVIDIA Vera Rubin) on Kimi K3 2.8T. 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
At 125 tok/s/user interactivity on Kimi K3 2.8T, B200 delivers 2217 tok/s/chip at $0.22 per million tokens; Vera Rubin NVL72 delivers 40202 tok/s/chip at $0.02. Vera Rubin NVL72 is 769% cheaper per token; Vera Rubin NVL72 delivers 1713% more tok/s/chip at this point.
B200 posts 1703 tok/s/chip for $0.28 per million tokens at 148 tok/s/user on Kimi K3 2.8T; Vera Rubin NVL72 posts 35774 tok/s/chip for $0.03. Vera Rubin NVL72 is 906% cheaper per token; Vera Rubin NVL72 delivers 2000% more tok/s/chip.
Throughput at 170 tok/s/user on Kimi K3 2.8T: B200 hits 1307 tok/s/chip, Vera Rubin NVL72 hits 31907. Per-million costs land at $0.37 and $0.03 respectively. Vera Rubin NVL72 is 1070% cheaper per token; Vera Rubin NVL72 delivers 2342% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:2216.9Vera Rubin NVL72:40201.7 | B200:1703.4Vera Rubin NVL72:35773.7 | B200:1306.8Vera Rubin NVL72:31907.2 |
| Cost ($/M tok) | B200:$0.217Vera Rubin NVL72:$0.025 | B200:$0.282Vera Rubin NVL72:$0.028 | B200:$0.368Vera Rubin NVL72:$0.031 |
| tok/s/MW | B200:1296406Vera Rubin NVL72:12182327 | B200:996165Vera Rubin NVL72:10840507 | B200:764183Vera Rubin NVL72:9668851 |
| Concurrency | B200:~6Vera Rubin NVL72:~224 | B200:~4Vera Rubin NVL72:~191 | B200:~3Vera Rubin NVL72:~162 |
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.
