Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and GB300 NVL72 (NVIDIA Blackwell) on Kimi K3 2.8T. 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 56 tok/s/user interactivity on Kimi K3 2.8T, GB200 NVL72 delivers 4146 tok/s/chip at $0.12 per million tokens; GB300 NVL72 delivers 12050 tok/s/chip at $0.05. GB300 NVL72 is 134% cheaper per token; GB300 NVL72 delivers 191% more tok/s/chip at this point.
GB200 NVL72 posts 2175 tok/s/chip for $0.24 per million tokens at 103 tok/s/user on Kimi K3 2.8T; GB300 NVL72 posts 7768 tok/s/chip for $0.08. GB300 NVL72 is 188% cheaper per token; GB300 NVL72 delivers 257% more tok/s/chip.
Throughput at 150 tok/s/user on Kimi K3 2.8T: GB200 NVL72 hits 1247 tok/s/chip, GB300 NVL72 hits 5640. Per-million costs land at $0.41 and $0.11 respectively. GB300 NVL72 is 264% cheaper per token; GB300 NVL72 delivers 352% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:4146.1GB300 NVL72:12050.0 | GB200 NVL72:2174.8GB300 NVL72:7767.6 | GB200 NVL72:1246.9GB300 NVL72:5640.0 |
| Cost ($/M tok) | GB200 NVL72:$0.125GB300 NVL72:$0.053 | GB200 NVL72:$0.238GB300 NVL72:$0.083 | GB200 NVL72:$0.414GB300 NVL72:$0.114 |
| tok/s/MW | GB200 NVL72:2217166GB300 NVL72:5683946 | GB200 NVL72:1162989GB300 NVL72:3663981 | GB200 NVL72:666816GB300 NVL72:2660380 |
| Concurrency | GB200 NVL72:~16GB300 NVL72:~48 | GB200 NVL72:~8GB300 NVL72:~32 | GB200 NVL72:~3GB300 NVL72:~27 |
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.
