Kimi K2.5/K2.6/K2.7-Code 1T — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) 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.
B200 / H200 on Kimi K2.5/K2.6/K2.7-Code 1T at 55 tok/s/user: 385 / 289 tok/s/GPU, $1.41 / $1.33 per million tokens. H200 is 6% cheaper per token; B200 delivers 33% more tok/s/GPU.
Around the middle of the 38–105 tok/s/user interactivity band, at 72 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: B200 runs 250 tok/s/GPU at $2.17/M tokens, H200 runs 217 at $1.79/M. H200 is 21% cheaper per token; B200 delivers 15% more tok/s/GPU.
Setting 89 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, B200 produces 168 tok/s/GPU ($3.21 per million tokens) and H200 produces 155 ($2.56). H200 is 26% cheaper per token; B200 delivers 9% more tok/s/GPU. (Numbers reflect the default 1k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:384.6H200:289.3 | B200:249.8H200:216.6 | B200:168.4H200:154.7 |
| Cost ($/M tok) | B200:$1.410H200:$1.328 | B200:$2.168H200:$1.795 | B200:$3.208H200:$2.556 |
| tok/s/MW | B200:177244H200:167223 | B200:115097H200:125187 | B200:77603H200:89400 |
| Concurrency | B200:~29H200:~21 | B200:~14H200:~12 | B200:~8H200:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.