Head-to-head AI inference benchmark comparison of B300 (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
Throughput at 70 tok/s/user on Kimi K3 2.8T: B300 hits 4851 tok/s/chip, TPU7x hits 5895. Per-million costs land at $0.13 and $0.05 respectively. TPU7x is 167% cheaper per token; TPU7x delivers 22% more tok/s/chip.
B300 / TPU7x on Kimi K3 2.8T at 114 tok/s/user: 3550 / 4094 tok/s/chip, $0.18 / $0.07 per million tokens. TPU7x is 153% cheaper per token; TPU7x delivers 15% more tok/s/chip.
Toward the upper edge of the 26β202 tok/s/user interactivity band, at 158 tok/s/user on Kimi K3 2.8T: B300 runs 2063 tok/s/chip at $0.30/M tokens, TPU7x runs 2447 at $0.12/M. TPU7x is 160% cheaper per token; TPU7x delivers 19% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:4851.0TPU7x:5895.0 | B300:3549.6TPU7x:4094.5 | B300:2063.2TPU7x:2446.8 |
| Cost ($/M tok) | B300:$0.129TPU7x:$0.049 | B300:$0.177TPU7x:$0.070 | B300:$0.304TPU7x:$0.117 |
| tok/s/MW | B300:2553136TPU7x:4883971 | B300:1868235TPU7x:3392285 | B300:1085910TPU7x:2027193 |
| Concurrency | B300:~7TPU7x:~39 | B300:~5TPU7x:~21 | B300:~1TPU7x:~11 |
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.
