gpt-oss 120B — MI300X vs MI325X
Head-to-head AI inference benchmark comparison of MI300X (AMD CDNA 3) 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.
Throughput at 50 tok/s/user on gpt-oss 120B: MI300X hits 3834 tok/s/GPU, MI325X hits 2737. Per-million costs land at $0.08 and $0.13 respectively. MI300X is 60% cheaper per token; MI300X delivers 40% more tok/s/GPU.
MI300X / MI325X on gpt-oss 120B at 71 tok/s/user: 3127 / 2084 tok/s/GPU, $0.10 / $0.17 per million tokens. MI300X is 67% cheaper per token; MI300X delivers 50% more tok/s/GPU.
Toward the upper edge of the 30–113 tok/s/user interactivity band, at 92 tok/s/user on gpt-oss 120B: MI300X runs 2051 tok/s/GPU at $0.15/M tokens, MI325X runs 1076 at $0.33/M. MI300X is 124% cheaper per token; MI300X delivers 91% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MI300X:3834.3MI325X:2737.2 | MI300X:3127.3MI325X:2084.1 | MI300X:2050.9MI325X:1075.9 |
| Cost ($/M tok) | MI300X:$0.081MI325X:$0.130 | MI300X:$0.102MI325X:$0.170 | MI300X:$0.148MI325X:$0.331 |
| tok/s/MW | MI300X:2142060MI325X:1255601 | MI300X:1747115MI325X:956001 | MI300X:1145772MI325X:493524 |
| Concurrency | MI300X:~64MI325X:~64 | MI300X:~49MI325X:~63 | MI300X:~23MI325X:~26 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.