MiniMax M3 428B — GB300 NVL72 vs MI300X
Head-to-head AI inference benchmark comparison of GB300 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.
Throughput at 45 tok/s/user on MiniMax M3 428B: GB300 NVL72 hits 1425 tok/s/GPU, MI300X hits 537. Per-million costs land at $0.49 and $0.57 respectively. GB300 NVL72 is 16% cheaper per token; GB300 NVL72 delivers 165% more tok/s/GPU.
GB300 NVL72 / MI300X on MiniMax M3 428B at 77 tok/s/user: 450 / 208 tok/s/GPU, $1.61 / $1.50 per million tokens. MI300X is 7% cheaper per token; GB300 NVL72 delivers 117% more tok/s/GPU.
Toward the upper edge of the 13–142 tok/s/user interactivity band, at 110 tok/s/user on MiniMax M3 428B: GB300 NVL72 runs 181 tok/s/GPU at $3.90/M tokens, MI300X runs 153 at $2.00/M. MI300X is 94% cheaper per token; GB300 NVL72 delivers 18% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB300 NVL72:1425.0MI300X:536.8 | GB300 NVL72:450.2MI300X:207.7 | GB300 NVL72:180.6MI300X:153.0 |
| Cost ($/M tok) | GB300 NVL72:$0.493MI300X:$0.570 | GB300 NVL72:$1.611MI300X:$1.502 | GB300 NVL72:$3.895MI300X:$2.003 |
| tok/s/MW | GB300 NVL72:678586MI300X:299909 | GB300 NVL72:214379MI300X:116006 | GB300 NVL72:85998MI300X:85462 |
| Concurrency | GB300 NVL72:~177MI300X:~51 | GB300 NVL72:~20MI300X:~12 | GB300 NVL72:~5MI300X:~6 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.