MiniMax M2.5/M2.7 — MI325X vs MI355X
Head-to-head AI inference benchmark comparison of MI325X (AMD CDNA 3) 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.
Near the low end of the 26–105 tok/s/user interactivity band, at 46 tok/s/user on MiniMax M2.5/M2.7: MI325X runs 1872 tok/s/GPU at $0.19/M tokens, MI355X runs 3584 at $0.12/M. MI355X is 64% cheaper per token; MI355X delivers 92% more tok/s/GPU.
Setting 66 tok/s/user as the target on MiniMax M2.5/M2.7, MI325X produces 857 tok/s/GPU ($0.41 per million tokens) and MI355X produces 1224 ($0.34). MI355X is 19% cheaper per token; MI355X delivers 43% more tok/s/GPU.
At 85 tok/s/user interactivity on MiniMax M2.5/M2.7, MI325X delivers 355 tok/s/GPU at $1.00 per million tokens; MI355X delivers 691 tok/s/GPU at $0.59. MI355X is 69% cheaper per token; MI355X delivers 95% more tok/s/GPU at this point. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MI325X:1871.6MI355X:3584.5 | MI325X:856.8MI355X:1224.1 | MI325X:354.6MI355X:691.4 |
| Cost ($/M tok) | MI325X:$0.190MI355X:$0.115 | MI325X:$0.407MI355X:$0.342 | MI325X:$1.003MI355X:$0.594 |
| tok/s/MW | MI325X:858544MI355X:1352629 | MI325X:393012MI355X:461908 | MI325X:162651MI355X:260920 |
| Concurrency | MI325X:~166MI355X:~256 | MI325X:~53MI355X:~39 | MI325X:~17MI355X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.