gpt-oss 120B — B200 vs MI325X
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) 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.
Setting 72 tok/s/user as the target on gpt-oss 120B, B200 produces 25320 tok/s/GPU ($0.02 per million tokens) and MI325X produces 1889 ($0.19). B200 is 782% cheaper per token; B200 delivers 1240% more tok/s/GPU.
At 86 tok/s/user interactivity on gpt-oss 120B, B200 delivers 22213 tok/s/GPU at $0.02 per million tokens; MI325X delivers 1111 tok/s/GPU at $0.31. B200 is 1181% cheaper per token; B200 delivers 1900% more tok/s/GPU at this point.
B200 posts 21001 tok/s/GPU for $0.03 per million tokens at 99 tok/s/user on gpt-oss 120B; MI325X posts 836 tok/s/GPU for $0.43. B200 is 1545% cheaper per token; B200 delivers 2412% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:25319.6MI325X:1888.9 | B200:22212.9MI325X:1110.9 | B200:21000.7MI325X:836.2 |
| Cost ($/M tok) | B200:$0.022MI325X:$0.190 | B200:$0.024MI325X:$0.312 | B200:$0.026MI325X:$0.426 |
| tok/s/MW | B200:11668029MI325X:866471 | B200:10236368MI325X:509578 | B200:9677720MI325X:383559 |
| Concurrency | B200:~256MI325X:~60 | B200:~256MI325X:~26 | B200:~251MI325X:~17 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.