Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) on GLM 5.3 744B. 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 146 tok/s/user on GLM 5.3 744B: B200 hits 8222 tok/s/chip, GB200 NVL72 hits 12180. Per-million costs land at $0.06 and $0.04 respectively. GB200 NVL72 is 38% cheaper per token; GB200 NVL72 delivers 48% more tok/s/chip.
B200 / GB200 NVL72 on GLM 5.3 744B at 197 tok/s/user: 6919 / 8252 tok/s/chip, $0.07 / $0.06 per million tokens. GB200 NVL72 is 11% cheaper per token; GB200 NVL72 delivers 19% more tok/s/chip.
Toward the upper edge of the 96–298 tok/s/user interactivity band, at 248 tok/s/user on GLM 5.3 744B: B200 runs 4416 tok/s/chip at $0.11/M tokens, GB200 NVL72 runs 4914 at $0.11/M. GB200 NVL72 is 4% cheaper per token; GB200 NVL72 delivers 11% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:8222.1GB200 NVL72:12179.7 | B200:6918.7GB200 NVL72:8252.5 | B200:4415.6GB200 NVL72:4914.2 |
| Cost ($/M tok) | B200:$0.058GB200 NVL72:$0.042 | B200:$0.069GB200 NVL72:$0.063 | B200:$0.109GB200 NVL72:$0.105 |
| tok/s/MW | B200:4808218GB200 NVL72:6513209 | B200:4046020GB200 NVL72:4413095 | B200:2582195GB200 NVL72:2627904 |
| Concurrency | B200:~12GB200 NVL72:~61 | B200:~8GB200 NVL72:~32 | B200:~5GB200 NVL72:~5 |
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.
