GLM 5/5.1 — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) on GLM 5/5.1. 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 GLM 5/5.1 at 41 tok/s/user: 458 / 329 tok/s/GPU, $1.19 / $1.19 per million tokens. Cost per token is essentially tied; B200 delivers 39% more tok/s/GPU.
Around the middle of the 21–101 tok/s/user interactivity band, at 61 tok/s/user on GLM 5/5.1: B200 runs 343 tok/s/GPU at $1.57/M tokens, H200 runs 212 at $1.85/M. B200 is 18% cheaper per token; B200 delivers 61% more tok/s/GPU.
Setting 82 tok/s/user as the target on GLM 5/5.1, B200 produces 209 tok/s/GPU ($2.62 per million tokens) and H200 produces 132 ($2.93). B200 is 12% cheaper per token; B200 delivers 57% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:458.3H200:329.5 | B200:342.9H200:212.5 | B200:208.5H200:132.4 |
| Cost ($/M tok) | B200:$1.188H200:$1.190 | B200:$1.569H200:$1.847 | B200:$2.623H200:$2.933 |
| tok/s/MW | B200:211190H200:190452 | B200:158024H200:122811 | B200:96102H200:76544 |
| Concurrency | B200:~473H200:~34 | B200:~22H200:~14 | B200:~11H200:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.