Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) 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
GB200 NVL72 / MI355X on DeepSeek V4.1 Flash 552B at 160 tok/s/user: 94237 / 85508 tok/s/chip, $0.01 / $0.00 per million tokens. MI355X is 13% cheaper per token; GB200 NVL72 delivers 10% more tok/s/chip.
Around the middle of the 43β515 tok/s/user interactivity band, at 279 tok/s/user on DeepSeek V4.1 Flash 552B: GB200 NVL72 runs 17037 tok/s/chip at $0.03/M tokens, MI355X runs 40939 at $0.01/M. MI355X is 198% cheaper per token; MI355X delivers 140% more tok/s/chip.
Setting 397 tok/s/user as the target on DeepSeek V4.1 Flash 552B, GB200 NVL72 produces 16012 tok/s/chip ($0.03 per million tokens) and MI355X produces 20774 ($0.02). MI355X is 61% cheaper per token; MI355X delivers 30% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:94237.2MI355X:85507.9 | GB200 NVL72:17036.6MI355X:40939.2 | GB200 NVL72:16012.5MI355X:20774.2 |
| Cost ($/M tok) | GB200 NVL72:$0.005MI355X:$0.005 | GB200 NVL72:$0.030MI355X:$0.010 | GB200 NVL72:$0.032MI355X:$0.020 |
| tok/s/MW | GB200 NVL72:50394205MI355X:40912857 | GB200 NVL72:9110465MI355X:19588150 | GB200 NVL72:8562823MI355X:9939822 |
| Concurrency | GB200 NVL72:~49MI355X:~32 | GB200 NVL72:~4MI355X:~11 | GB200 NVL72:~8MI355X:~4 |
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.
