MiniMax M3 428B — H100 vs MI325X
Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) 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.
Setting 57 tok/s/user as the target on MiniMax M3 428B, H100 produces 776 tok/s/GPU ($0.47 per million tokens) and MI325X produces 485 ($0.73). H100 is 57% cheaper per token; H100 delivers 60% more tok/s/GPU.
At 98 tok/s/user interactivity on MiniMax M3 428B, H100 delivers 358 tok/s/GPU at $1.01 per million tokens; MI325X delivers 200 tok/s/GPU at $1.78. H100 is 76% cheaper per token; H100 delivers 79% more tok/s/GPU at this point.
H100 posts 232 tok/s/GPU for $1.54 per million tokens at 140 tok/s/user on MiniMax M3 428B; MI325X posts 113 tok/s/GPU for $3.13. H100 is 103% cheaper per token; H100 delivers 105% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:776.1MI325X:485.4 | H100:358.5MI325X:200.1 | H100:232.4MI325X:113.3 |
| Cost ($/M tok) | H100:$0.467MI325X:$0.734 | H100:$1.008MI325X:$1.777 | H100:$1.540MI325X:$3.128 |
| tok/s/MW | H100:448586MI325X:222651 | H100:207213MI325X:91808 | H100:134316MI325X:51985 |
| Concurrency | H100:~59MI325X:~38 | H100:~16MI325X:~9 | H100:~7MI325X:~3 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.