MiniMax M3 428B — B200 vs B300
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and B300 (NVIDIA Blackwell) on MiniMax M3 428B. 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.
B200 posts 855 tok/s/GPU for $0.63 per million tokens at 59 tok/s/user on MiniMax M3 428B; B300 posts 1664 tok/s/GPU for $0.39. B300 is 62% cheaper per token; B300 delivers 95% more tok/s/GPU.
Throughput at 106 tok/s/user on MiniMax M3 428B: B200 hits 334 tok/s/GPU, B300 hits 463. Per-million costs land at $1.61 and $1.42 respectively. B300 is 14% cheaper per token; B300 delivers 39% more tok/s/GPU.
B200 / B300 on MiniMax M3 428B at 152 tok/s/user: 125 / 225 tok/s/GPU, $4.32 / $2.83 per million tokens. B300 is 52% cheaper per token; B300 delivers 80% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:855.3B300:1663.8 | B200:333.7B300:463.4 | B200:125.5B300:225.5 |
| Cost ($/M tok) | B200:$0.633B300:$0.392 | B200:$1.614B300:$1.421 | B200:$4.317B300:$2.833 |
| tok/s/MW | B200:394152B300:766735 | B200:153771B300:213530 | B200:57823B300:103902 |
| Concurrency | B200:~32B300:~63 | B200:~8B300:~20 | B200:~4B300:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.