DeepSeek V4.1 Flash 552B — GB300 NVL72 vs MI300X
Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and MI300X (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 — 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 →
Near the low end of the 44–205 tok/s/user interactivity band, at 84 tok/s/user on DeepSeek V4.1 Flash 552B: GB300 NVL72 runs 1094347 tok/s/chip at $0.00/M tokens, MI300X runs 9358 at $0.03/M. GB300 NVL72 is 4709% cheaper per token; GB300 NVL72 delivers 11594% more tok/s/chip.
Setting 124 tok/s/user as the target on DeepSeek V4.1 Flash 552B, GB300 NVL72 produces 1040710 tok/s/chip ($0.00 per million tokens) and MI300X produces 6243 ($0.04). GB300 NVL72 is 6756% cheaper per token; GB300 NVL72 delivers 16570% more tok/s/chip.
At 165 tok/s/user interactivity on DeepSeek V4.1 Flash 552B, GB300 NVL72 delivers 976288 tok/s/chip at $0.00 per million tokens; MI300X delivers 3943 tok/s/chip at $0.07. GB300 NVL72 is 10083% cheaper per token; GB300 NVL72 delivers 24660% more tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:1094347.1MI300X:9358.4 | GB300 NVL72:1040710.1MI300X:6242.9 | GB300 NVL72:976288.2MI300X:3943.0 |
| Cost ($/M tok) | GB300 NVL72:$0.001MI300X:$0.028 | GB300 NVL72:$0.001MI300X:$0.042 | GB300 NVL72:$0.001MI300X:$0.067 |
| tok/s/MW | GB300 NVL72:516201439MI300X:6732631 | GB300 NVL72:490900996MI300X:4491265 | GB300 NVL72:460513316MI300X:2836683 |
| Concurrency | GB300 NVL72:~7480MI300X:~7 | GB300 NVL72:~7480MI300X:~4 | GB300 NVL72:~8591MI300X:~1 |
Inference Performance
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.