MiniMax M2.5/M2.7 — H100 vs MI300X
Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and MI300X (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.
H100 / MI300X on MiniMax M2.5/M2.7 at 53 tok/s/user: 741 / 994 tok/s/GPU, $0.48 / $0.31 per million tokens. MI300X is 55% cheaper per token; MI300X delivers 34% more tok/s/GPU.
Around the middle of the 40–95 tok/s/user interactivity band, at 67 tok/s/user on MiniMax M2.5/M2.7: H100 runs 497 tok/s/GPU at $0.72/M tokens, MI300X runs 627 at $0.50/M. MI300X is 46% cheaper per token; MI300X delivers 26% more tok/s/GPU.
Setting 82 tok/s/user as the target on MiniMax M2.5/M2.7, H100 produces 344 tok/s/GPU ($1.04 per million tokens) and MI300X produces 351 ($0.88). MI300X is 19% cheaper per token; MI300X delivers 2% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:740.9MI300X:994.1 | H100:497.2MI300X:627.0 | H100:343.5MI300X:350.7 |
| Cost ($/M tok) | H100:$0.483MI300X:$0.311 | H100:$0.724MI300X:$0.495 | H100:$1.044MI300X:$0.881 |
| tok/s/MW | H100:428240MI300X:555352 | H100:287416MI300X:350284 | H100:198569MI300X:195927 |
| Concurrency | H100:~57MI300X:~41 | H100:~30MI300X:~19 | H100:~8MI300X:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.