MiniMax M2.5/M2.7 — B200 vs GB200 NVL72
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) 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.
Setting 68 tok/s/user as the target on MiniMax M2.5/M2.7, B200 produces 8180 tok/s/chip ($0.06 per million tokens) and GB200 NVL72 produces 8433 ($0.06). B200 is 4% cheaper per token; GB200 NVL72 delivers 3% more tok/s/chip.
At 105 tok/s/user interactivity on MiniMax M2.5/M2.7, B200 delivers 2975 tok/s/chip at $0.16 per million tokens; GB200 NVL72 delivers 3247 tok/s/chip at $0.16. Cost per token is essentially tied; GB200 NVL72 delivers 9% more tok/s/chip at this point.
B200 posts 1143 tok/s/chip for $0.42 per million tokens at 143 tok/s/user on MiniMax M2.5/M2.7; GB200 NVL72 posts 1066 tok/s/chip for $0.49. B200 is 16% cheaper per token; B200 delivers 7% more tok/s/chip. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:8180.3GB200 NVL72:8432.8 | B200:2975.5GB200 NVL72:3246.9 | B200:1143.3GB200 NVL72:1066.4 |
| Cost ($/M tok) | B200:$0.059GB200 NVL72:$0.061 | B200:$0.162GB200 NVL72:$0.160 | B200:$0.420GB200 NVL72:$0.489 |
| tok/s/MW | B200:4783779GB200 NVL72:4509515 | B200:1740053GB200 NVL72:1736305 | B200:668598GB200 NVL72:570241 |
| Concurrency | B200:~831GB200 NVL72:~928 | B200:~106GB200 NVL72:~236 | B200:~18GB200 NVL72:~45 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.