Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB200 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
Throughput at 209 tok/s/user on DeepSeek V4.1 Flash 552B: B200 hits 156778 tok/s/chip, GB200 NVL72 hits 154577. Per-million costs land at $0.00 and $0.00 respectively. B200 is 9% cheaper per token; B200 delivers 1% more tok/s/chip.
B200 / GB200 NVL72 on DeepSeek V4.1 Flash 552B at 332 tok/s/user: 63634 / 61313 tok/s/chip, $0.01 / $0.01 per million tokens. B200 is 12% cheaper per token; B200 delivers 4% more tok/s/chip.
Toward the upper edge of the 86β578 tok/s/user interactivity band, at 455 tok/s/user on DeepSeek V4.1 Flash 552B: B200 runs 22322 tok/s/chip at $0.02/M tokens, GB200 NVL72 runs 14327 at $0.04/M. B200 is 68% cheaper per token; B200 delivers 56% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:156778.2GB200 NVL72:154577.3 | B200:63633.6GB200 NVL72:61313.2 | B200:22322.4GB200 NVL72:14326.7 |
| Cost ($/M tok) | B200:$0.003GB200 NVL72:$0.003 | B200:$0.008GB200 NVL72:$0.008 | B200:$0.022GB200 NVL72:$0.036 |
| tok/s/MW | B200:91683150GB200 NVL72:82661682 | B200:37212660GB200 NVL72:32787792 | B200:13054016GB200 NVL72:7661349 |
| Concurrency | B200:~93GB200 NVL72:~92 | B200:~35GB200 NVL72:~34 | B200:~12GB200 NVL72:~7 |
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.
