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: 468 / 329 tok/s/GPU, $1.16 / $1.19 per million tokens. B200 is 3% cheaper per token; B200 delivers 42% 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 392 tok/s/GPU at $1.39/M tokens, H200 runs 212 at $1.85/M. B200 is 33% cheaper per token; B200 delivers 85% more tok/s/GPU.
Setting 82 tok/s/user as the target on GLM 5/5.1, B200 produces 275 tok/s/GPU ($1.97 per million tokens) and H200 produces 132 ($2.93). B200 is 49% cheaper per token; B200 delivers 108% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:468.2H200:329.5 | B200:392.4H200:212.5 | B200:275.4H200:132.4 |
| Cost ($/M tok) | B200:$1.159H200:$1.190 | B200:$1.391H200:$1.847 | B200:$1.972H200:$2.933 |
| tok/s/MW | B200:215746H200:190452 | B200:180829H200:122811 | B200:126911H200:76544 |
| Concurrency | B200:~484H200:~34 | B200:~25H200:~14 | B200:~14H200:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.