Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and Vera Rubin NVL72 (NVIDIA Vera Rubin) on DeepSeekv4 Pro 0813 1.6T. 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 →
At 101 tok/s/user interactivity on DeepSeekv4 Pro 0813 1.6T, GB200 NVL72 delivers 4168 tok/s/chip at $0.12 per million tokens; Vera Rubin NVL72 delivers 195187 tok/s/chip at $0.01. Vera Rubin NVL72 is 2313% cheaper per token; Vera Rubin NVL72 delivers 4583% more tok/s/chip at this point.
GB200 NVL72 posts 3276 tok/s/chip for $0.16 per million tokens at 127 tok/s/user on DeepSeekv4 Pro 0813 1.6T; Vera Rubin NVL72 posts 167288 tok/s/chip for $0.01. Vera Rubin NVL72 is 2531% cheaper per token; Vera Rubin NVL72 delivers 5006% more tok/s/chip.
Throughput at 154 tok/s/user on DeepSeekv4 Pro 0813 1.6T: GB200 NVL72 hits 2438 tok/s/chip, Vera Rubin NVL72 hits 111699. Per-million costs land at $0.21 and $0.01 respectively. Vera Rubin NVL72 is 2261% cheaper per token; Vera Rubin NVL72 delivers 4482% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:4167.9Vera Rubin NVL72:195187.2 | GB200 NVL72:3276.3Vera Rubin NVL72:167288.0 | GB200 NVL72:2437.8Vera Rubin NVL72:111699.0 |
| Cost ($/M tok) | GB200 NVL72:$0.124Vera Rubin NVL72:$0.005 | GB200 NVL72:$0.158Vera Rubin NVL72:$0.006 | GB200 NVL72:$0.212Vera Rubin NVL72:$0.009 |
| tok/s/MW | GB200 NVL72:2228815Vera Rubin NVL72:59147627 | GB200 NVL72:1752014Vera Rubin NVL72:50693345 | GB200 NVL72:1303649Vera Rubin NVL72:33848177 |
| Concurrency | GB200 NVL72:~6Vera Rubin NVL72:~1775 | GB200 NVL72:~5Vera Rubin NVL72:~1137 | GB200 NVL72:~3Vera Rubin NVL72:~749 |
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
DeepSeek V4 Pro 0813 1.6T 8K / 1K Total Tokens per $1 TCO vs. Interactivity