gpt-oss 120B · GPU comparison

gpt-oss 120B — H100 vs MI300X

Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) 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.

Setting 112 tok/s/user as the target on gpt-oss 120B, H100 produces 2796 tok/s/GPU ($0.13 per million tokens) and MI300X produces 1599 ($0.20). H100 is 50% cheaper per token; H100 delivers 75% more tok/s/GPU.

At 158 tok/s/user interactivity on gpt-oss 120B, H100 delivers 1521 tok/s/GPU at $0.24 per million tokens; MI300X delivers 803 tok/s/GPU at $0.37. H100 is 57% cheaper per token; H100 delivers 90% more tok/s/GPU at this point.

H100 posts 878 tok/s/GPU for $0.41 per million tokens at 203 tok/s/user on gpt-oss 120B; MI300X posts 309 tok/s/GPU for $0.97. H100 is 138% cheaper per token; H100 delivers 184% 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.)

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)
H100:2795.7MI300X:1598.8
H100:1521.3MI300X:802.8
H100:878.3MI300X:309.1
Cost ($/M tok)
H100:$0.131MI300X:$0.195
H100:$0.238MI300X:$0.373
H100:$0.409MI300X:$0.975
tok/s/MW
H100:1616027MI300X:893159
H100:879363MI300X:448476
H100:507706MI300X:172687
Concurrency
H100:~64MI300X:~15
H100:~28MI300X:~5
H100:~10MI300X:~7

Inference Performance

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