MiniMax M2.5/M2.7 — H200 vs MI355X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI355X (AMD CDNA 4) 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.
Throughput at 48 tok/s/user on MiniMax M2.5/M2.7: H200 hits 1288 tok/s/GPU, MI355X hits 3518. Per-million costs land at $0.31 and $0.12 respectively. MI355X is 165% cheaper per token; MI355X delivers 173% more tok/s/GPU.
H200 / MI355X on MiniMax M2.5/M2.7 at 72 tok/s/user: 538 / 1047 tok/s/GPU, $0.73 / $0.39 per million tokens. MI355X is 86% cheaper per token; MI355X delivers 95% more tok/s/GPU.
Toward the upper edge of the 25–119 tok/s/user interactivity band, at 96 tok/s/user on MiniMax M2.5/M2.7: H200 runs 389 tok/s/GPU at $1.00/M tokens, MI355X runs 528 at $0.75/M. MI355X is 34% cheaper per token; MI355X delivers 36% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:1287.5MI355X:3518.5 | H200:537.6MI355X:1046.7 | H200:389.3MI355X:527.8 |
| Cost ($/M tok) | H200:$0.313MI355X:$0.118 | H200:$0.733MI355X:$0.393 | H200:$1.000MI355X:$0.747 |
| tok/s/MW | H200:744221MI355X:1327732 | H200:310779MI355X:394984 | H200:225034MI355X:199177 |
| Concurrency | H200:~64MI355X:~256 | H200:~30MI355X:~31 | H200:~9MI355X:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.