MiniMax M2.5/M2.7 — H200 vs MI300X
Head-to-head AI inference benchmark comparison of H200 (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.
Near the low end of the 39–95 tok/s/user interactivity band, at 53 tok/s/user on MiniMax M2.5/M2.7: H200 runs 1159 tok/s/GPU at $0.33/M tokens, MI300X runs 994 at $0.31/M. MI300X is 6% cheaper per token; H200 delivers 17% more tok/s/GPU.
Setting 67 tok/s/user as the target on MiniMax M2.5/M2.7, H200 produces 662 tok/s/GPU ($0.61 per million tokens) and MI300X produces 627 ($0.50). MI300X is 23% cheaper per token; H200 delivers 6% more tok/s/GPU.
At 81 tok/s/user interactivity on MiniMax M2.5/M2.7, H200 delivers 410 tok/s/GPU at $0.92 per million tokens; MI300X delivers 365 tok/s/GPU at $0.84. MI300X is 10% cheaper per token; H200 delivers 12% more tok/s/GPU at this point. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:1158.8MI300X:994.1 | H200:661.9MI300X:627.0 | H200:410.4MI300X:365.5 |
| Cost ($/M tok) | H200:$0.331MI300X:$0.311 | H200:$0.607MI300X:$0.495 | H200:$0.924MI300X:$0.843 |
| tok/s/MW | H200:669855MI300X:555352 | H200:382573MI300X:350284 | H200:237241MI300X:204170 |
| Concurrency | H200:~91MI300X:~41 | H200:~41MI300X:~19 | H200:~21MI300X:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.