Llama 3.3 70B — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) on Llama 3.3 70B. 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 Llama 3.3 70B at 65 tok/s/user: 5366 / 2459 tok/s/GPU, $0.10 / $0.16 per million tokens. B200 is 55% cheaper per token; B200 delivers 118% more tok/s/GPU.
Around the middle of the 34–159 tok/s/user interactivity band, at 97 tok/s/user on Llama 3.3 70B: B200 runs 3284 tok/s/GPU at $0.16/M tokens, H200 runs 1279 at $0.31/M. B200 is 89% cheaper per token; B200 delivers 157% more tok/s/GPU.
Setting 128 tok/s/user as the target on Llama 3.3 70B, B200 produces 1649 tok/s/GPU ($0.33 per million tokens) and H200 produces 522 ($0.75). B200 is 125% cheaper per token; B200 delivers 216% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:5365.8H200:2458.7 | B200:3284.3H200:1278.7 | B200:1648.9H200:521.5 |
| Cost ($/M tok) | B200:$0.102H200:$0.158 | B200:$0.163H200:$0.309 | B200:$0.333H200:$0.751 |
| tok/s/MW | B200:2472729H200:1421206 | B200:1513525H200:739144 | B200:759842H200:301461 |
| Concurrency | B200:~128H200:~64 | B200:~74H200:~27 | B200:~27H200:~16 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.