gpt-oss 120B — GB200 NVL72 vs MI300X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) 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 85 tok/s/user interactivity on gpt-oss 120B, GB200 NVL72 delivers 40563 tok/s/GPU at $0.02 per million tokens; MI300X delivers 2267 tok/s/GPU at $0.14. GB200 NVL72 is 792% cheaper per token; GB200 NVL72 delivers 1689% more tok/s/GPU at this point.
GB200 NVL72 posts 18202 tok/s/GPU for $0.03 per million tokens at 140 tok/s/user on gpt-oss 120B; MI300X posts 1014 tok/s/GPU for $0.31. GB200 NVL72 is 804% cheaper per token; GB200 NVL72 delivers 1695% more tok/s/GPU.
Throughput at 194 tok/s/user on gpt-oss 120B: GB200 NVL72 hits 11222 tok/s/GPU, MI300X hits 383. Per-million costs land at $0.06 and $0.81 respectively. GB200 NVL72 is 1357% cheaper per token; GB200 NVL72 delivers 2830% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:40563.4MI300X:2267.0 | GB200 NVL72:18201.5MI300X:1014.2 | GB200 NVL72:11222.5MI300X:383.0 |
| Cost ($/M tok) | GB200 NVL72:$0.015MI300X:$0.135 | GB200 NVL72:$0.034MI300X:$0.305 | GB200 NVL72:$0.056MI300X:$0.814 |
| tok/s/MW | GB200 NVL72:19315900MI300X:1266489 | GB200 NVL72:8667398MI300X:566576 | GB200 NVL72:5344037MI300X:213994 |
| Concurrency | GB200 NVL72:~2217MI300X:~27 | GB200 NVL72:~2076MI300X:~7 | GB200 NVL72:~436MI300X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.