MiniMax M2.5/M2.7 — H100 vs MI355X
Head-to-head AI inference benchmark comparison of H100 (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.
H100 posts 599 tok/s/chip for $0.54 per million tokens at 60 tok/s/user on MiniMax M2.5/M2.7; MI355X posts 1691 tok/s/chip for $0.25. MI355X is 120% cheaper per token; MI355X delivers 182% more tok/s/chip.
Throughput at 79 tok/s/user on MiniMax M2.5/M2.7: H100 hits 353 tok/s/chip, MI355X hits 856. Per-million costs land at $0.92 and $0.49 respectively. MI355X is 89% cheaper per token; MI355X delivers 143% more tok/s/chip.
H100 / MI355X on MiniMax M2.5/M2.7 at 98 tok/s/user: 207 / 503 tok/s/chip, $1.57 / $0.83 per million tokens. MI355X is 90% cheaper per token; MI355X delivers 143% more tok/s/chip. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H100:599.4MI355X:1691.0 | H100:353.0MI355X:856.2 | H100:207.1MI355X:503.3 |
| Cost ($/M tok) | H100:$0.542MI355X:$0.246 | H100:$0.921MI355X:$0.487 | H100:$1.569MI355X:$0.828 |
| tok/s/MW | H100:437502MI355X:809091 | H100:257637MI355X:409662 | H100:151188MI355X:240807 |
| Concurrency | H100:~41MI355X:~46 | H100:~18MI355X:~20 | H100:~9MI355X:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.