MiniMax M2.5/M2.7 — B300 vs MI325X
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) 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.
At 46 tok/s/user interactivity on MiniMax M2.5/M2.7, B300 delivers 5248 tok/s/GPU at $0.12 per million tokens; MI325X delivers 1872 tok/s/GPU at $0.19. B300 is 53% cheaper per token; B300 delivers 180% more tok/s/GPU at this point.
B300 posts 2710 tok/s/GPU for $0.24 per million tokens at 66 tok/s/user on MiniMax M2.5/M2.7; MI325X posts 857 tok/s/GPU for $0.41. B300 is 71% cheaper per token; B300 delivers 216% more tok/s/GPU.
Throughput at 85 tok/s/user on MiniMax M2.5/M2.7: B300 hits 1403 tok/s/GPU, MI325X hits 355. Per-million costs land at $0.47 and $1.00 respectively. B300 is 115% cheaper per token; B300 delivers 296% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:5248.4MI325X:1871.6 | B300:2709.9MI325X:856.8 | B300:1402.7MI325X:354.6 |
| Cost ($/M tok) | B300:$0.124MI325X:$0.190 | B300:$0.238MI325X:$0.407 | B300:$0.466MI325X:$1.003 |
| tok/s/MW | B300:2418637MI325X:858544 | B300:1248784MI325X:393012 | B300:646404MI325X:162651 |
| Concurrency | B300:~234MI325X:~166 | B300:~84MI325X:~53 | B300:~34MI325X:~17 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.