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 117 tok/s/user: 2622 / 8080 tok/s/GPU, $0.14 / $0.05 per million tokens. MI355X is 172% cheaper per token; MI355X delivers 208% more tok/s/GPU.
Around the middle of the 67–266 tok/s/user interactivity band, at 166 tok/s/user on gpt-oss 120B: H100 runs 1379 tok/s/GPU at $0.26/M tokens, MI355X runs 2826 at $0.15/M. MI355X is 76% cheaper per token; MI355X delivers 105% more tok/s/GPU.
Setting 216 tok/s/user as the target on gpt-oss 120B, H100 produces 740 tok/s/GPU ($0.49 per million tokens) and MI355X produces 1784 ($0.23). MI355X is 112% cheaper per token; MI355X delivers 141% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:2621.5MI355X:8080.1 | H100:1379.3MI355X:2825.6 | H100:739.7MI355X:1783.6 |
| Cost ($/M tok) | H100:$0.139MI355X:$0.051 | H100:$0.262MI355X:$0.149 | H100:$0.487MI355X:$0.230 |
| tok/s/MW | H100:1515329MI355X:3049088 | H100:797303MI355X:1066268 | H100:427578MI355X:673045 |
| Concurrency | H100:~64MI355X:~36 | H100:~17MI355X:~30 | H100:~8MI355X:~31 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.