gpt-oss 120B — H100 vs MI355X
Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) 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.
H100 / MI355X on gpt-oss 120B at 119 tok/s/user: 2552 / 8968 tok/s/GPU, $0.14 / $0.05 per million tokens. MI355X is 214% cheaper per token; MI355X delivers 251% more tok/s/GPU.
Around the middle of the 71–266 tok/s/user interactivity band, at 168 tok/s/user on gpt-oss 120B: H100 runs 1347 tok/s/GPU at $0.27/M tokens, MI355X runs 4252 at $0.10/M. MI355X is 179% cheaper per token; MI355X delivers 216% more tok/s/GPU.
Setting 217 tok/s/user as the target on gpt-oss 120B, H100 produces 730 tok/s/GPU ($0.49 per million tokens) and MI355X produces 2312 ($0.18). MI355X is 178% cheaper per token; MI355X delivers 217% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:2552.2MI355X:8967.9 | H100:1347.2MI355X:4251.9 | H100:730.4MI355X:2312.2 |
| Cost ($/M tok) | H100:$0.143MI355X:$0.046 | H100:$0.268MI355X:$0.096 | H100:$0.492MI355X:$0.177 |
| tok/s/MW | H100:1475239MI355X:3384126 | H100:778709MI355X:1604507 | H100:422184MI355X:872516 |
| Concurrency | H100:~64MI355X:~39 | H100:~16MI355X:~13 | H100:~8MI355X:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.