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 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 138 tok/s/user on GLM 5.3 744B: B200 hits 10091 tok/s/chip, GB200 NVL72 hits 14039. Per-million costs land at $0.05 and $0.04 respectively. GB200 NVL72 is 29% cheaper per token; GB200 NVL72 delivers 39% more tok/s/chip.
B200 / GB200 NVL72 on GLM 5.3 744B at 192 tok/s/user: 6951 / 8833 tok/s/chip, $0.07 / $0.06 per million tokens. GB200 NVL72 is 18% cheaper per token; GB200 NVL72 delivers 27% more tok/s/chip.
Toward the upper edge of the 84β301 tok/s/user interactivity band, at 247 tok/s/user on GLM 5.3 744B: B200 runs 4541 tok/s/chip at $0.11/M tokens, GB200 NVL72 runs 4779 at $0.11/M. B200 is 2% cheaper per token; GB200 NVL72 delivers 5% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:10090.6GB200 NVL72:14038.8 | B200:6950.8GB200 NVL72:8833.2 | B200:4541.2GB200 NVL72:4779.3 |
| Cost ($/M tok) | B200:$0.048GB200 NVL72:$0.037 | B200:$0.069GB200 NVL72:$0.058 | B200:$0.106GB200 NVL72:$0.108 |
| tok/s/MW | B200:5900942GB200 NVL72:7507402 | B200:4064818GB200 NVL72:4723651 | B200:2655655GB200 NVL72:2555793 |
| Concurrency | B200:~34GB200 NVL72:~74 | B200:~8GB200 NVL72:~45 | 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.
