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 40–104 tok/s/user interactivity band, at 56 tok/s/user on MiniMax M2.5/M2.7: H100 runs 669 tok/s/GPU at $0.53/M tokens, MI325X runs 1108 at $0.32/M. MI325X is 66% cheaper per token; MI325X delivers 66% more tok/s/GPU.
Setting 72 tok/s/user as the target on MiniMax M2.5/M2.7, H100 produces 435 tok/s/GPU ($0.81 per million tokens) and MI325X produces 576 ($0.62). MI325X is 32% cheaper per token; MI325X delivers 32% more tok/s/GPU.
At 88 tok/s/user interactivity on MiniMax M2.5/M2.7, H100 delivers 285 tok/s/GPU at $1.30 per million tokens; MI325X delivers 283 tok/s/GPU at $1.23. MI325X is 6% cheaper per token; throughput per GPU is essentially tied at this point. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:668.8MI325X:1108.3 | H100:435.4MI325X:575.7 | H100:284.9MI325X:282.9 |
| Cost ($/M tok) | H100:$0.531MI325X:$0.320 | H100:$0.815MI325X:$0.619 | H100:$1.302MI325X:$1.228 |
| tok/s/MW | H100:386568MI325X:508386 | H100:251684MI325X:264090 | H100:164667MI325X:129774 |
| Concurrency | H100:~48MI325X:~83 | H100:~21MI325X:~9 | H100:~8MI325X:~13 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.