Kimi K3 2.8T · Chip comparison
Kimi K3 2.8T — B300 vs TPU7x
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 — 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 →
Throughput at 70 tok/s/user on Kimi K3 2.8T: B300 hits 4,851 tok/s/chip, TPU7x hits 5,895. Costs per million tokens land at $0.13 and $0.049 respectively. TPU7x is 62% cheaper per token; TPU7x delivers 22% more tok/s/chip.
B300 / TPU7x on Kimi K3 2.8T at 114 tok/s/user: 3,550 / 4,094 tok/s/chip, $0.18 / $0.070 per million tokens. TPU7x is 60% cheaper per token; TPU7x delivers 15% more tok/s/chip.
Toward the upper edge of the 26–202 tok/s/user interactivity band, at 158 tok/s/user on Kimi K3 2.8T: B300 runs 2,063 tok/s/chip at $0.30 per million tokens, TPU7x runs 2,447 at $0.12. TPU7x is 62% cheaper per token; TPU7x delivers 19% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B3004,851TPU7x5,895 | B3003,550TPU7x4,094 | B3002,063TPU7x2,447 |
| Cost ($/M tok) | B300$0.13TPU7x$0.049 | B300$0.18TPU7x$0.070 | B300$0.30TPU7x$0.12 |
| tok/s/MW | B3002,553,136TPU7x4,883,971 | B3001,868,235TPU7x3,392,285 | B3001,085,910TPU7x2,027,193 |
| Concurrency | B300~7TPU7x~39 | B300~5TPU7x~21 | B300~1TPU7x~11 |