Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) 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 →
B200 posts 4452 tok/s/chip for $0.11 per million tokens at 52 tok/s/user on Kimi K3 2.8T; H200 posts 70 tok/s/chip for $4.85. B200 is 4393% cheaper per token; B200 delivers 6271% more tok/s/chip.
Throughput at 104 tok/s/user on Kimi K3 2.8T: B200 hits 2102 tok/s/chip, H200 hits 70. Per-million costs land at $0.23 and $4.85 respectively. B200 is 2022% cheaper per token; B200 delivers 2908% more tok/s/chip.
B200 / H200 on Kimi K3 2.8T at 155 tok/s/user: 1271 / 70 tok/s/chip, $0.38 / $4.85 per million tokens. B200 is 1183% cheaper per token; B200 delivers 1719% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:4451.9H200:69.9 | B200:2102.0H200:69.9 | B200:1271.0H200:69.9 |
| Cost ($/M tok) | B200:$0.108H200:$4.850 | B200:$0.229H200:$4.850 | B200:$0.378H200:$4.850 |
| tok/s/MW | B200:2603424H200:51001 | B200:1229257H200:51001 | B200:743296H200:51001 |
| Concurrency | B200:~16H200:~1 | B200:~6H200:~1 | B200:~3H200:~1 |
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.
