gpt-oss 120B · GPU comparison

gpt-oss 120B — GB200 NVL72 vs MI355X

Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) 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.

Near the low end of the 63–304 tok/s/user interactivity band, at 123 tok/s/user on gpt-oss 120B: GB200 NVL72 runs 18583 tok/s/GPU at $0.03/M tokens, MI355X runs 7338 at $0.06/M. GB200 NVL72 is 69% cheaper per token; GB200 NVL72 delivers 153% more tok/s/GPU.

Setting 183 tok/s/user as the target on gpt-oss 120B, GB200 NVL72 produces 13317 tok/s/GPU ($0.05 per million tokens) and MI355X produces 2493 ($0.17). GB200 NVL72 is 246% cheaper per token; GB200 NVL72 delivers 434% more tok/s/GPU.

At 244 tok/s/user interactivity on gpt-oss 120B, GB200 NVL72 delivers 6601 tok/s/GPU at $0.09 per million tokens; MI355X delivers 1252 tok/s/GPU at $0.34. GB200 NVL72 is 269% cheaper per token; GB200 NVL72 delivers 427% more tok/s/GPU at this point. (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.)

View performance-per-dollar view →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/gpu)
GB200 NVL72:18583.2MI355X:7338.3
GB200 NVL72:13317.5MI355X:2493.0
GB200 NVL72:6600.6MI355X:1252.0
Cost ($/M tok)
GB200 NVL72:$0.033MI355X:$0.056
GB200 NVL72:$0.048MI355X:$0.165
GB200 NVL72:$0.091MI355X:$0.337
tok/s/MW
GB200 NVL72:8849142MI355X:2769176
GB200 NVL72:6341659MI355X:940755
GB200 NVL72:3143146MI355X:472436
Concurrency
GB200 NVL72:~406MI355X:~31
GB200 NVL72:~1272MI355X:~9
GB200 NVL72:~71MI355X:~27

Inference Performance

Inference performance metrics across different models, hardware configurations, and serving parameters.