MiniMax M3 428B — GB200 NVL72 vs MI300X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI300X (AMD CDNA 3) on MiniMax M3 428B. 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.
GB200 NVL72 posts 1803 tok/s/GPU for $0.34 per million tokens at 50 tok/s/user on MiniMax M3 428B; MI300X posts 492 tok/s/GPU for $0.62. GB200 NVL72 is 83% cheaper per token; GB200 NVL72 delivers 266% more tok/s/GPU.
Throughput at 83 tok/s/user on MiniMax M3 428B: GB200 NVL72 hits 712 tok/s/GPU, MI300X hits 200. Per-million costs land at $0.89 and $1.55 respectively. GB200 NVL72 is 73% cheaper per token; GB200 NVL72 delivers 257% more tok/s/GPU.
GB200 NVL72 / MI300X on MiniMax M3 428B at 116 tok/s/user: 227 / 140 tok/s/GPU, $2.66 / $2.18 per million tokens. MI300X is 22% cheaper per token; GB200 NVL72 delivers 62% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:1802.8MI300X:492.2 | GB200 NVL72:712.4MI300X:199.7 | GB200 NVL72:227.1MI300X:140.2 |
| Cost ($/M tok) | GB200 NVL72:$0.340MI300X:$0.622 | GB200 NVL72:$0.895MI300X:$1.549 | GB200 NVL72:$2.665MI300X:$2.178 |
| tok/s/MW | GB200 NVL72:858484MI300X:274963 | GB200 NVL72:339243MI300X:111582 | GB200 NVL72:108150MI300X:78332 |
| Concurrency | GB200 NVL72:~217MI300X:~42 | GB200 NVL72:~83MI300X:~10 | GB200 NVL72:~20MI300X:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.