Kimi K2.5/K2.6/K2.7-Code 1T — GB200 NVL72 vs GB300 NVL72
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and GB300 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.
Near the low end of the 18–186 tok/s/user interactivity band, at 60 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: GB200 NVL72 runs 7627 tok/s/chip at $0.07/M tokens, GB300 NVL72 runs 7910 at $0.08/M. GB200 NVL72 is 18% cheaper per token; GB300 NVL72 delivers 4% more tok/s/chip.
Setting 102 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, GB200 NVL72 produces 451 tok/s/chip ($1.18 per million tokens) and GB300 NVL72 produces 621 ($1.04). GB300 NVL72 is 13% cheaper per token; GB300 NVL72 delivers 38% more tok/s/chip.
At 144 tok/s/user interactivity on Kimi K2.5/K2.6/K2.7-Code 1T, GB200 NVL72 delivers 192 tok/s/chip at $2.65 per million tokens; GB300 NVL72 delivers 212 tok/s/chip at $3.01. GB200 NVL72 is 14% cheaper per token; GB300 NVL72 delivers 10% more tok/s/chip at this point. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:7626.9GB300 NVL72:7910.3 | GB200 NVL72:451.0GB300 NVL72:621.5 | GB200 NVL72:192.1GB300 NVL72:211.9 |
| Cost ($/M tok) | GB200 NVL72:$0.068GB300 NVL72:$0.080 | GB200 NVL72:$1.184GB300 NVL72:$1.045 | GB200 NVL72:$2.652GB300 NVL72:$3.014 |
| tok/s/MW | GB200 NVL72:4078558GB300 NVL72:3731285 | GB200 NVL72:241164GB300 NVL72:293151 | GB200 NVL72:102727GB300 NVL72:99942 |
| Concurrency | GB200 NVL72:~3340GB300 NVL72:~3377 | GB200 NVL72:~98GB300 NVL72:~135 | GB200 NVL72:~28GB300 NVL72:~29 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.