Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and TPU7x (Google Ironwood) 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
GB200 NVL72 posts 3432 tok/s/chip for $0.15 per million tokens at 68 tok/s/user on Kimi K3 2.8T; TPU7x posts 5981 tok/s/chip for $0.05. TPU7x is 215% cheaper per token; TPU7x delivers 74% more tok/s/chip.
Throughput at 112 tok/s/user on Kimi K3 2.8T: GB200 NVL72 hits 2050 tok/s/chip, TPU7x hits 4175. Per-million costs land at $0.25 and $0.07 respectively. TPU7x is 268% cheaper per token; TPU7x delivers 104% more tok/s/chip.
GB200 NVL72 / TPU7x on Kimi K3 2.8T at 155 tok/s/user: 963 / 2554 tok/s/chip, $0.54 / $0.11 per million tokens. TPU7x is 379% cheaper per token; TPU7x delivers 165% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:3431.5TPU7x:5981.4 | GB200 NVL72:2050.0TPU7x:4175.0 | GB200 NVL72:962.5TPU7x:2554.4 |
| Cost ($/M tok) | GB200 NVL72:$0.151TPU7x:$0.048 | GB200 NVL72:$0.252TPU7x:$0.069 | GB200 NVL72:$0.537TPU7x:$0.112 |
| tok/s/MW | GB200 NVL72:1835033TPU7x:4955578 | GB200 NVL72:1096256TPU7x:3459016 | GB200 NVL72:514709TPU7x:2116310 |
| Concurrency | GB200 NVL72:~16TPU7x:~40 | GB200 NVL72:~14TPU7x:~22 | GB200 NVL72:~2TPU7x:~12 |
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.
