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 46 tok/s/user as the target on MiniMax M2.5/M2.7, H200 produces 1407 tok/s/GPU ($0.28 per million tokens) and MI325X produces 1643 ($0.22). MI325X is 29% cheaper per token; MI325X delivers 17% more tok/s/GPU.
At 65 tok/s/user interactivity on MiniMax M2.5/M2.7, H200 delivers 718 tok/s/GPU at $0.56 per million tokens; MI325X delivers 757 tok/s/GPU at $0.48. MI325X is 16% cheaper per token; MI325X delivers 5% more tok/s/GPU at this point.
H200 posts 443 tok/s/GPU for $0.87 per million tokens at 85 tok/s/user on MiniMax M2.5/M2.7; MI325X posts 325 tok/s/GPU for $1.09. H200 is 26% cheaper per token; H200 delivers 36% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:1407.2MI325X:1642.6 | H200:717.8MI325X:757.2 | H200:442.8MI325X:324.7 |
| Cost ($/M tok) | H200:$0.281MI325X:$0.217 | H200:$0.556MI325X:$0.479 | H200:$0.868MI325X:$1.094 |
| tok/s/MW | H200:813437MI325X:753479 | H200:414916MI325X:347355 | H200:255930MI325X:148934 |
| Concurrency | H200:~64MI325X:~147 | H200:~51MI325X:~38 | H200:~15MI325X:~16 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.