Kimi K2.5/K2.6/K2.7-Code 1T — B300 vs GB200 NVL72
Head-to-head AI inference benchmark comparison of B300 (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.
At 69 tok/s/user interactivity on Kimi K2.5/K2.6/K2.7-Code 1T, B300 delivers 932 tok/s/GPU at $0.70 per million tokens; GB200 NVL72 delivers 5669 tok/s/GPU at $0.11. GB200 NVL72 is 547% cheaper per token; GB200 NVL72 delivers 509% more tok/s/GPU at this point.
B300 posts 899 tok/s/GPU for $0.72 per million tokens at 104 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T; GB200 NVL72 posts 435 tok/s/GPU for $1.45. B300 is 100% cheaper per token; B300 delivers 107% more tok/s/GPU.
Throughput at 138 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: B300 hits 493 tok/s/GPU, GB200 NVL72 hits 220. Per-million costs land at $1.32 and $2.79 respectively. B300 is 111% cheaper per token; B300 delivers 124% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:931.5GB200 NVL72:5669.1 | B300:899.5GB200 NVL72:434.6 | B300:493.4GB200 NVL72:220.2 |
| Cost ($/M tok) | B300:$0.698GB200 NVL72:$0.108 | B300:$0.723GB200 NVL72:$1.446 | B300:$1.323GB200 NVL72:$2.789 |
| tok/s/MW | B300:429274GB200 NVL72:2699555 | B300:414503GB200 NVL72:206965 | B300:227387GB200 NVL72:104845 |
| Concurrency | B300:~28GB200 NVL72:~1936 | B300:~19GB200 NVL72:~89 | B300:~7GB200 NVL72:~30 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.