gpt-oss 120B — H200 vs MI300X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI300X (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.
At 96 tok/s/user interactivity on gpt-oss 120B, H200 delivers 3494 tok/s/GPU at $0.11 per million tokens; MI300X delivers 1956 tok/s/GPU at $0.16. H200 is 39% cheaper per token; H200 delivers 79% more tok/s/GPU at this point.
H200 posts 1980 tok/s/GPU for $0.20 per million tokens at 147 tok/s/user on gpt-oss 120B; MI300X posts 919 tok/s/GPU for $0.33. H200 is 67% cheaper per token; H200 delivers 116% more tok/s/GPU.
Throughput at 198 tok/s/user on gpt-oss 120B: H200 hits 1037 tok/s/GPU, MI300X hits 346. Per-million costs land at $0.37 and $0.89 respectively. H200 is 139% cheaper per token; H200 delivers 199% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:3494.1MI300X:1956.0 | H200:1980.4MI300X:918.9 | H200:1037.0MI300X:346.5 |
| Cost ($/M tok) | H200:$0.112MI300X:$0.155 | H200:$0.198MI300X:$0.330 | H200:$0.373MI300X:$0.890 |
| tok/s/MW | H200:2019715MI300X:1092764 | H200:1144761MI300X:513352 | H200:599403MI300X:193555 |
| Concurrency | H200:~64MI300X:~21 | H200:~62MI300X:~6 | H200:~16MI300X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.