Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) 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
At 61 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, GB200 NVL72 delivers 175618 tok/s/chip at $0.00 per million tokens; MI325X delivers 15608 tok/s/chip at $0.02. GB200 NVL72 is 565% cheaper per token; GB200 NVL72 delivers 1025% more tok/s/chip at this point.
GB200 NVL72 posts 140378 tok/s/chip for $0.00 per million tokens at 112 tok/s/user on DeepSeek V4.1 Flash 552B; MI325X posts 9279 tok/s/chip for $0.03. GB200 NVL72 is 795% cheaper per token; GB200 NVL72 delivers 1413% more tok/s/chip.
Throughput at 164 tok/s/user on DeepSeek V4.1 Flash 552B: GB200 NVL72 hits 91807 tok/s/chip, MI325X hits 4124. Per-million costs land at $0.01 and $0.07 respectively. GB200 NVL72 is 1216% cheaper per token; GB200 NVL72 delivers 2126% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:175618.3MI325X:15608.2 | GB200 NVL72:140378.4MI325X:9279.5 | GB200 NVL72:91806.7MI325X:4124.2 |
| Cost ($/M tok) | GB200 NVL72:$0.003MI325X:$0.020 | GB200 NVL72:$0.004MI325X:$0.033 | GB200 NVL72:$0.006MI325X:$0.074 |
| tok/s/MW | GB200 NVL72:93913545MI325X:9235634 | GB200 NVL72:75068641MI325X:5490810 | GB200 NVL72:49094511MI325X:2440359 |
| Concurrency | GB200 NVL72:~128MI325X:~6 | GB200 NVL72:~86MI325X:~2 | GB200 NVL72:~48MI325X:~1 |
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.
