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 5020 tok/s/GPU at $0.13 per million tokens; MI325X delivers 1643 tok/s/GPU at $0.22. B300 is 68% cheaper per token; B300 delivers 206% more tok/s/GPU at this point.
B300 posts 2770 tok/s/GPU for $0.24 per million tokens at 65 tok/s/user on MiniMax M2.5/M2.7; MI325X posts 757 tok/s/GPU for $0.48. B300 is 100% cheaper per token; B300 delivers 266% more tok/s/GPU.
Throughput at 85 tok/s/user on MiniMax M2.5/M2.7: B300 hits 1380 tok/s/GPU, MI325X hits 325. Per-million costs land at $0.47 and $1.09 respectively. B300 is 132% cheaper per token; B300 delivers 325% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:5020.0MI325X:1642.6 | B300:2770.2MI325X:757.2 | B300:1379.5MI325X:324.7 |
| Cost ($/M tok) | B300:$0.129MI325X:$0.217 | B300:$0.240MI325X:$0.479 | B300:$0.472MI325X:$1.094 |
| tok/s/MW | B300:2313350MI325X:753479 | B300:1276587MI325X:347355 | B300:635730MI325X:148934 |
| Concurrency | B300:~455MI325X:~147 | B300:~101MI325X:~38 | B300:~34MI325X:~16 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.