gpt-oss 120B — B200 vs H100
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H100 (NVIDIA Hopper) on gpt-oss 120B. 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.
Throughput at 117 tok/s/user on gpt-oss 120B: B200 hits 20258 tok/s/GPU, H100 hits 2622. Per-million costs land at $0.03 and $0.14 respectively. B200 is 417% cheaper per token; B200 delivers 673% more tok/s/GPU.
B200 / H100 on gpt-oss 120B at 166 tok/s/user: 15502 / 1379 tok/s/GPU, $0.04 / $0.26 per million tokens. B200 is 632% cheaper per token; B200 delivers 1024% more tok/s/GPU.
Toward the upper edge of the 67–266 tok/s/user interactivity band, at 216 tok/s/user on gpt-oss 120B: B200 runs 8506 tok/s/GPU at $0.06/M tokens, H100 runs 740 at $0.49/M. B200 is 700% cheaper per token; B200 delivers 1050% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:20257.7H100:2621.5 | B200:15502.5H100:1379.3 | B200:8505.6H100:739.7 |
| Cost ($/M tok) | B200:$0.027H100:$0.139 | B200:$0.036H100:$0.262 | B200:$0.061H100:$0.487 |
| tok/s/MW | B200:9335363H100:1515329 | B200:7144005H100:797303 | B200:3919610H100:427578 |
| Concurrency | B200:~219H100:~64 | B200:~92H100:~17 | B200:~64H100:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.