Head-to-head AI inference benchmark comparison of B200 (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
Setting 67 tok/s/user as the target on Kimi K3 2.8T, B200 produces 4436 tok/s/chip ($0.11 per million tokens) and TPU7x produces 6023 ($0.05). TPU7x is 128% cheaper per token; TPU7x delivers 36% more tok/s/chip.
At 109 tok/s/user interactivity on Kimi K3 2.8T, B200 delivers 2664 tok/s/chip at $0.18 per million tokens; TPU7x delivers 4296 tok/s/chip at $0.07. TPU7x is 171% cheaper per token; TPU7x delivers 61% more tok/s/chip at this point.
B200 posts 1643 tok/s/chip for $0.29 per million tokens at 151 tok/s/user on Kimi K3 2.8T; TPU7x posts 2702 tok/s/chip for $0.11. TPU7x is 176% cheaper per token; TPU7x delivers 64% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:4436.1TPU7x:6023.1 | B200:2663.8TPU7x:4296.3 | B200:1642.9TPU7x:2701.6 |
| Cost ($/M tok) | B200:$0.108TPU7x:$0.048 | B200:$0.180TPU7x:$0.067 | B200:$0.293TPU7x:$0.106 |
| tok/s/MW | B200:2594227TPU7x:4990167 | B200:1557766TPU7x:3559521 | B200:960737TPU7x:2238281 |
| Concurrency | B200:~15TPU7x:~40 | B200:~7TPU7x:~23 | B200:~4TPU7x:~13 |
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.
