MiniMax M3 428B — B300 vs GB300 NVL72
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB300 NVL72 (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.
Throughput at 42 tok/s/user on MiniMax M3 428B: B300 hits 3498 tok/s/GPU, GB300 NVL72 hits 1607. Per-million costs land at $0.19 and $0.42 respectively. B300 is 127% cheaper per token; B300 delivers 118% more tok/s/GPU.
B300 / GB300 NVL72 on MiniMax M3 428B at 76 tok/s/user: 1595 / 458 tok/s/GPU, $0.42 / $1.58 per million tokens. B300 is 280% cheaper per token; B300 delivers 248% more tok/s/GPU.
Toward the upper edge of the 9–142 tok/s/user interactivity band, at 109 tok/s/user on MiniMax M3 428B: B300 runs 834 tok/s/GPU at $0.78/M tokens, GB300 NVL72 runs 185 at $3.79/M. B300 is 387% cheaper per token; B300 delivers 350% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:3497.8GB300 NVL72:1607.2 | B300:1595.3GB300 NVL72:458.1 | B300:834.5GB300 NVL72:185.3 |
| Cost ($/M tok) | B300:$0.186GB300 NVL72:$0.423 | B300:$0.416GB300 NVL72:$1.579 | B300:$0.779GB300 NVL72:$3.791 |
| tok/s/MW | B300:1611900GB300 NVL72:765340 | B300:735151GB300 NVL72:218131 | B300:384547GB300 NVL72:88227 |
| Concurrency | B300:~183GB300 NVL72:~225 | B300:~64GB300 NVL72:~21 | B300:~38GB300 NVL72:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.