gpt-oss 120B — H200 vs MI325X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI325X (AMD CDNA 3) 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.
H200 posts 4718 tok/s/GPU for $0.08 per million tokens at 58 tok/s/user on gpt-oss 120B; MI325X posts 2494 tok/s/GPU for $0.14. H200 is 71% cheaper per token; H200 delivers 89% more tok/s/GPU.
Throughput at 76 tok/s/user on gpt-oss 120B: H200 hits 4201 tok/s/GPU, MI325X hits 1618. Per-million costs land at $0.09 and $0.23 respectively. H200 is 141% cheaper per token; H200 delivers 160% more tok/s/GPU.
H200 / MI325X on gpt-oss 120B at 94 tok/s/user: 3563 / 948 tok/s/GPU, $0.11 / $0.37 per million tokens. H200 is 235% cheaper per token; H200 delivers 276% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:4718.1MI325X:2493.7 | H200:4200.9MI325X:1618.4 | H200:3562.9MI325X:947.7 |
| Cost ($/M tok) | H200:$0.083MI325X:$0.141 | H200:$0.093MI325X:$0.225 | H200:$0.110MI325X:$0.367 |
| tok/s/MW | H200:2727236MI325X:1143916 | H200:2428280MI325X:742376 | H200:2059472MI325X:434721 |
| Concurrency | H200:~64MI325X:~64 | H200:~64MI325X:~49 | H200:~64MI325X:~21 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.