gpt-oss 120B — B200 vs MI300X
Head-to-head AI inference benchmark comparison of B200 (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 106 tok/s/user interactivity on gpt-oss 120B, B200 delivers 18505 tok/s/chip at $0.03 per million tokens; MI300X delivers 1744 tok/s/chip at $0.15. B200 is 484% cheaper per token; B200 delivers 961% more tok/s/chip at this point.
B200 posts 10626 tok/s/chip for $0.05 per million tokens at 154 tok/s/user on gpt-oss 120B; MI300X posts 842 tok/s/chip for $0.30. B200 is 562% cheaper per token; B200 delivers 1162% more tok/s/chip.
Throughput at 201 tok/s/user on gpt-oss 120B: B200 hits 6666 tok/s/chip, MI300X hits 323. Per-million costs land at $0.07 and $0.80 respectively. B200 is 1012% cheaper per token; B200 delivers 1963% more tok/s/chip. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:18505.3MI300X:1743.9 | B200:10625.6MI300X:842.2 | B200:6665.6MI300X:323.1 |
| Cost ($/M tok) | B200:$0.026MI300X:$0.151 | B200:$0.046MI300X:$0.301 | B200:$0.072MI300X:$0.798 |
| tok/s/MW | B200:10821819MI300X:1254619 | B200:6213799MI300X:605925 | B200:3897997MI300X:232449 |
| Concurrency | B200:~207MI300X:~17 | B200:~66MI300X:~5 | B200:~64MI300X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.