MiniMax M2.5/M2.7 — H100 vs MI325X
Head-to-head AI inference benchmark comparison of H100 (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.
Near the low end of the 40–105 tok/s/user interactivity band, at 56 tok/s/user on MiniMax M2.5/M2.7: H100 runs 693 tok/s/GPU at $0.52/M tokens, MI325X runs 1283 at $0.28/M. MI325X is 85% cheaper per token; MI325X delivers 85% more tok/s/GPU.
Setting 72 tok/s/user as the target on MiniMax M2.5/M2.7, H100 produces 430 tok/s/GPU ($0.82 per million tokens) and MI325X produces 697 ($0.49). MI325X is 67% cheaper per token; MI325X delivers 62% more tok/s/GPU.
At 89 tok/s/user interactivity on MiniMax M2.5/M2.7, H100 delivers 276 tok/s/GPU at $1.34 per million tokens; MI325X delivers 278 tok/s/GPU at $1.27. MI325X is 5% cheaper per token; throughput per GPU is essentially tied at this point. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:693.0MI325X:1283.0 | H100:429.5MI325X:696.8 | H100:276.4MI325X:277.9 |
| Cost ($/M tok) | H100:$0.517MI325X:$0.279 | H100:$0.820MI325X:$0.491 | H100:$1.340MI325X:$1.270 |
| tok/s/MW | H100:400569MI325X:588543 | H100:248271MI325X:319654 | H100:159793MI325X:127466 |
| Concurrency | H100:~50MI325X:~94 | H100:~20MI325X:~40 | H100:~8MI325X:~13 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.