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 1272 tok/s/chip, MI355X hits 1854. Per-million costs land at $0.28 and $0.22 respectively. MI355X is 23% cheaper per token; MI355X delivers 46% more tok/s/chip.
H200 / MI355X on MiniMax M2.5/M2.7 at 72 tok/s/user: 641 / 1047 tok/s/chip, $0.50 / $0.40 per million tokens. MI355X is 25% cheaper per token; MI355X delivers 63% more tok/s/chip.
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 394 tok/s/chip at $0.86/M tokens, MI355X runs 528 at $0.76/M. MI355X is 13% cheaper per token; MI355X delivers 34% more tok/s/chip. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H200:1272.3MI355X:1853.7 | H200:640.5MI355X:1046.7 | H200:393.8MI355X:527.8 |
| Cost ($/M tok) | H200:$0.276MI355X:$0.225 | H200:$0.499MI355X:$0.398 | H200:$0.856MI355X:$0.757 |
| tok/s/MW | H200:928660MI355X:886922 | H200:467522MI355X:500816 | H200:287436MI355X:252545 |
| Concurrency | H200:~54MI355X:~511 | H200:~19MI355X:~31 | H200:~9MI355X:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.