MiniMax M2.5/M2.7 — GB200 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) 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 42 tok/s/user on MiniMax M2.5/M2.7: GB200 NVL72 hits 6025 tok/s/chip, MI355X hits 3180. Per-million costs land at $0.09 and $0.13 respectively. GB200 NVL72 is 56% cheaper per token; GB200 NVL72 delivers 89% more tok/s/chip.
GB200 NVL72 / MI355X on MiniMax M2.5/M2.7 at 68 tok/s/user: 2643 / 1185 tok/s/chip, $0.20 / $0.36 per million tokens. GB200 NVL72 is 76% cheaper per token; GB200 NVL72 delivers 123% more tok/s/chip.
Toward the upper edge of the 16–119 tok/s/user interactivity band, at 93 tok/s/user on MiniMax M2.5/M2.7: GB200 NVL72 runs 790 tok/s/chip at $0.67/M tokens, MI355X runs 568 at $0.71/M. GB200 NVL72 is 7% cheaper per token; GB200 NVL72 delivers 39% more tok/s/chip. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:6024.6MI355X:3180.5 | GB200 NVL72:2642.8MI355X:1184.6 | GB200 NVL72:789.9MI355X:567.7 |
| Cost ($/M tok) | GB200 NVL72:$0.086MI355X:$0.133 | GB200 NVL72:$0.202MI355X:$0.355 | GB200 NVL72:$0.668MI355X:$0.712 |
| tok/s/MW | GB200 NVL72:3221721MI355X:1521763 | GB200 NVL72:1413237MI355X:566770 | GB200 NVL72:422432MI355X:271640 |
| Concurrency | GB200 NVL72:~1024MI355X:~141 | GB200 NVL72:~209MI355X:~17 | GB200 NVL72:~32MI355X:~6 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.