Qwen 3.5 397B-A17B — H200 vs MI325X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI325X (AMD CDNA 3) on Qwen 3.5 397B-A17B. 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.
H200 / MI325X on Qwen 3.5 397B-A17B at 46 tok/s/user: 600 / 403 tok/s/GPU, $0.65 / $0.86 per million tokens. H200 is 34% cheaper per token; H200 delivers 49% more tok/s/GPU.
Around the middle of the 37–72 tok/s/user interactivity band, at 55 tok/s/user on Qwen 3.5 397B-A17B: H200 runs 549 tok/s/GPU at $0.71/M tokens, MI325X runs 219 at $1.63/M. H200 is 131% cheaper per token; H200 delivers 150% more tok/s/GPU.
Setting 64 tok/s/user as the target on Qwen 3.5 397B-A17B, H200 produces 510 tok/s/GPU ($0.77 per million tokens) and MI325X produces 134 ($2.64). H200 is 244% cheaper per token; H200 delivers 280% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:600.2MI325X:402.9 | H200:549.4MI325X:219.4 | H200:510.5MI325X:134.3 |
| Cost ($/M tok) | H200:$0.646MI325X:$0.864 | H200:$0.706MI325X:$1.631 | H200:$0.767MI325X:$2.638 |
| tok/s/MW | H200:346909MI325X:184830 | H200:317594MI325X:100633 | H200:295059MI325X:61591 |
| Concurrency | H200:~52MI325X:~37 | H200:~40MI325X:~16 | H200:~32MI325X:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.