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 / 2960 tok/s/GPU, $0.03 / $0.13 per million tokens. B200 is 397% cheaper per token; B200 delivers 589% more tok/s/GPU.
Around the middle of the 59–273 tok/s/user interactivity band, at 166 tok/s/user on gpt-oss 120B: B200 runs 15458 tok/s/GPU at $0.04/M tokens, H200 runs 1473 at $0.27/M. B200 is 644% cheaper per token; B200 delivers 949% more tok/s/GPU.
Setting 220 tok/s/user as the target on gpt-oss 120B, B200 produces 8473 tok/s/GPU ($0.06 per million tokens) and H200 produces 747 ($0.52). B200 is 733% cheaper per token; B200 delivers 1035% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:20405.5H200:2959.9 | B200:15458.2H200:1473.4 | B200:8473.4H200:746.7 |
| Cost ($/M tok) | B200:$0.027H200:$0.133 | B200:$0.036H200:$0.268 | B200:$0.062H200:$0.520 |
| tok/s/MW | B200:9403440H200:1710947 | B200:7123584H200:851677 | B200:3904812H200:431599 |
| Concurrency | B200:~230H200:~64 | B200:~92H200:~23 | B200:~64H200:~14 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.