gpt-oss 120B — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (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.
B200 / H200 on gpt-oss 120B at 112 tok/s/user: 20405 / 2969 tok/s/GPU, $0.03 / $0.13 per million tokens. B200 is 395% cheaper per token; B200 delivers 587% more tok/s/GPU.
Around the middle of the 59–270 tok/s/user interactivity band, at 164 tok/s/user on gpt-oss 120B: B200 runs 16066 tok/s/GPU at $0.03/M tokens, H200 runs 1607 at $0.24/M. B200 is 603% cheaper per token; B200 delivers 900% more tok/s/GPU.
Setting 217 tok/s/user as the target on gpt-oss 120B, B200 produces 8662 tok/s/GPU ($0.06 per million tokens) and H200 produces 797 ($0.49). B200 is 707% cheaper per token; B200 delivers 986% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:20405.5H200:2968.7 | B200:16065.7H200:1607.1 | B200:8662.1H200:797.4 |
| Cost ($/M tok) | B200:$0.027H200:$0.132 | B200:$0.035H200:$0.243 | B200:$0.061H200:$0.493 |
| tok/s/MW | B200:9403440H200:1715993 | B200:7403526H200:928953 | B200:3991748H200:460922 |
| Concurrency | B200:~230H200:~64 | B200:~98H200:~45 | B200:~64H200:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.