MiniMax M2.5/M2.7 — H200 vs MI325X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI325X (AMD CDNA 3) on MiniMax M2.5/M2.7. 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 55 tok/s/user as the target on MiniMax M2.5/M2.7, H200 produces 1088 tok/s/GPU ($0.35 per million tokens) and MI325X produces 1336 ($0.27). MI325X is 31% cheaper per token; MI325X delivers 23% more tok/s/GPU.
At 72 tok/s/user interactivity on MiniMax M2.5/M2.7, H200 delivers 527 tok/s/GPU at $0.74 per million tokens; MI325X delivers 697 tok/s/GPU at $0.49. MI325X is 50% cheaper per token; MI325X delivers 32% more tok/s/GPU at this point.
H200 posts 356 tok/s/GPU for $1.10 per million tokens at 88 tok/s/user on MiniMax M2.5/M2.7; MI325X posts 295 tok/s/GPU for $1.20. H200 is 9% cheaper per token; H200 delivers 21% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:1087.8MI325X:1335.6 | H200:527.5MI325X:696.8 | H200:355.7MI325X:294.8 |
| Cost ($/M tok) | H200:$0.352MI325X:$0.268 | H200:$0.737MI325X:$0.491 | H200:$1.099MI325X:$1.201 |
| tok/s/MW | H200:628771MI325X:612661 | H200:304895MI325X:319654 | H200:205598MI325X:135248 |
| Concurrency | H200:~82MI325X:~99 | H200:~30MI325X:~40 | H200:~17MI325X:~14 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.