Kimi K2.5/K2.6/K2.7-Code 1T — B300 vs GB300 NVL72
Head-to-head AI inference benchmark comparison of B300 (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.
B300 / GB300 NVL72 on Kimi K2.5/K2.6/K2.7-Code 1T at 69 tok/s/user: 932 / 5779 tok/s/chip, $0.67 / $0.11 per million tokens. GB300 NVL72 is 508% cheaper per token; GB300 NVL72 delivers 520% more tok/s/chip.
Around the middle of the 35–172 tok/s/user interactivity band, at 104 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: B300 runs 899 tok/s/chip at $0.70/M tokens, GB300 NVL72 runs 576 at $1.12/M. B300 is 60% cheaper per token; B300 delivers 56% more tok/s/chip.
Setting 138 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, B300 produces 493 tok/s/chip ($1.28 per million tokens) and GB300 NVL72 produces 223 ($2.88). B300 is 125% cheaper per token; B300 delivers 121% more tok/s/chip. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:931.5GB300 NVL72:5779.4 | B300:899.5GB300 NVL72:576.5 | B300:493.4GB300 NVL72:222.9 |
| Cost ($/M tok) | B300:$0.675GB300 NVL72:$0.111 | B300:$0.699GB300 NVL72:$1.119 | B300:$1.278GB300 NVL72:$2.879 |
| tok/s/MW | B300:490276GB300 NVL72:2726124 | B300:473406GB300 NVL72:271928 | B300:259700GB300 NVL72:105140 |
| Concurrency | B300:~28GB300 NVL72:~1977 | B300:~19GB300 NVL72:~119 | B300:~7GB300 NVL72:~30 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.