MiniMax M3 428B — B300 vs MI325X
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI325X (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.
B300 posts 1887 tok/s/GPU for $0.34 per million tokens at 55 tok/s/user on MiniMax M3 428B; MI325X posts 513 tok/s/GPU for $0.69. B300 is 106% cheaper per token; B300 delivers 268% more tok/s/GPU.
Throughput at 97 tok/s/user on MiniMax M3 428B: B300 hits 581 tok/s/GPU, MI325X hits 203. Per-million costs land at $1.13 and $1.75 respectively. B300 is 55% cheaper per token; B300 delivers 186% more tok/s/GPU.
B300 / MI325X on MiniMax M3 428B at 140 tok/s/user: 255 / 113 tok/s/GPU, $2.54 / $3.13 per million tokens. B300 is 23% cheaper per token; B300 delivers 125% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:1887.4MI325X:512.9 | B300:581.5MI325X:203.0 | B300:254.8MI325X:113.3 |
| Cost ($/M tok) | B300:$0.338MI325X:$0.695 | B300:$1.131MI325X:$1.750 | B300:$2.543MI325X:$3.128 |
| tok/s/MW | B300:869786MI325X:235281 | B300:267968MI325X:93105 | B300:117411MI325X:51985 |
| Concurrency | B300:~76MI325X:~41 | B300:~29MI325X:~9 | B300:~8MI325X:~3 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.