Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and GB300 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
GB200 NVL72 / GB300 NVL72 on GLM 5.3 744B at 142 tok/s/user: 12490 / 15450 tok/s/chip, $0.04 / $0.04 per million tokens. Cost per token is essentially tied; GB300 NVL72 delivers 24% more tok/s/chip.
Around the middle of the 92β291 tok/s/user interactivity band, at 192 tok/s/user on GLM 5.3 744B: GB200 NVL72 runs 8895 tok/s/chip at $0.06/M tokens, GB300 NVL72 runs 7697 at $0.08/M. GB200 NVL72 is 44% cheaper per token; GB200 NVL72 delivers 16% more tok/s/chip.
Setting 242 tok/s/user as the target on GLM 5.3 744B, GB200 NVL72 produces 5110 tok/s/chip ($0.10 per million tokens) and GB300 NVL72 produces 4849 ($0.13). GB200 NVL72 is 31% cheaper per token; GB200 NVL72 delivers 5% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:12490.1GB300 NVL72:15450.0 | GB200 NVL72:8895.0GB300 NVL72:7697.1 | GB200 NVL72:5110.2GB300 NVL72:4849.2 |
| Cost ($/M tok) | GB200 NVL72:$0.041GB300 NVL72:$0.042 | GB200 NVL72:$0.058GB300 NVL72:$0.083 | GB200 NVL72:$0.101GB300 NVL72:$0.132 |
| tok/s/MW | GB200 NVL72:6679177GB300 NVL72:7287742 | GB200 NVL72:4756708GB300 NVL72:3630696 | GB200 NVL72:2732702GB300 NVL72:2287336 |
| Concurrency | GB200 NVL72:~64GB300 NVL72:~48 | GB200 NVL72:~45GB300 NVL72:~43 | GB200 NVL72:~5GB300 NVL72:~8 |
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.
