Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and GB300 NVL72 (NVIDIA Blackwell) on DeepSeek V4.1 Flash 552B. 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
Setting 209 tok/s/user as the target on DeepSeek V4.1 Flash 552B, GB200 NVL72 produces 154577 tok/s/chip ($0.00 per million tokens) and GB300 NVL72 produces 932864 ($0.00). GB300 NVL72 is 386% cheaper per token; GB300 NVL72 delivers 503% more tok/s/chip.
At 332 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, GB200 NVL72 delivers 61313 tok/s/chip at $0.01 per million tokens; GB300 NVL72 delivers 455992 tok/s/chip at $0.00. GB300 NVL72 is 499% cheaper per token; GB300 NVL72 delivers 644% more tok/s/chip at this point.
GB200 NVL72 posts 14327 tok/s/chip for $0.04 per million tokens at 455 tok/s/user on DeepSeek V4.1 Flash 552B; GB300 NVL72 posts 21118 tok/s/chip for $0.03. GB300 NVL72 is 19% cheaper per token; GB300 NVL72 delivers 47% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:154577.3GB300 NVL72:932863.6 | GB200 NVL72:61313.2GB300 NVL72:455991.9 | GB200 NVL72:14326.7GB300 NVL72:21118.2 |
| Cost ($/M tok) | GB200 NVL72:$0.003GB300 NVL72:$0.001 | GB200 NVL72:$0.008GB300 NVL72:$0.001 | GB200 NVL72:$0.036GB300 NVL72:$0.030 |
| tok/s/MW | GB200 NVL72:82661682GB300 NVL72:440030023 | GB200 NVL72:32787792GB300 NVL72:215090510 | GB200 NVL72:7661349GB300 NVL72:9961417 |
| Concurrency | GB200 NVL72:~92GB300 NVL72:~10753 | GB200 NVL72:~34GB300 NVL72:~2970 | GB200 NVL72:~7GB300 NVL72:~11 |
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.
