Kimi K2.5/K2.6/K2.7-Code 1T — B200 vs GB200 NVL72
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) on Kimi K2.5/K2.6/K2.7-Code 1T. 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 Kimi K2.5/K2.6/K2.7-Code 1T, B200 produces 785 tok/s/GPU ($0.68 per million tokens) and GB200 NVL72 produces 5984 ($0.10). GB200 NVL72 is 569% cheaper per token; GB200 NVL72 delivers 662% more tok/s/GPU.
At 102 tok/s/user interactivity on Kimi K2.5/K2.6/K2.7-Code 1T, B200 delivers 415 tok/s/GPU at $1.30 per million tokens; GB200 NVL72 delivers 451 tok/s/GPU at $1.41. B200 is 8% cheaper per token; GB200 NVL72 delivers 9% more tok/s/GPU at this point.
B200 posts 207 tok/s/GPU for $2.73 per million tokens at 136 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T; GB200 NVL72 posts 230 tok/s/GPU for $2.66. GB200 NVL72 is 3% cheaper per token; GB200 NVL72 delivers 11% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:784.9GB200 NVL72:5983.5 | B200:414.9GB200 NVL72:451.0 | B200:206.7GB200 NVL72:230.1 |
| Cost ($/M tok) | B200:$0.684GB200 NVL72:$0.102 | B200:$1.303GB200 NVL72:$1.407 | B200:$2.733GB200 NVL72:$2.657 |
| tok/s/MW | B200:361718GB200 NVL72:2849292 | B200:191190GB200 NVL72:214751 | B200:95276GB200 NVL72:109581 |
| Concurrency | B200:~24GB200 NVL72:~2093 | B200:~8GB200 NVL72:~98 | B200:~3GB200 NVL72:~31 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.