DeepSeek V4 Pro 1.6T — H200 vs MI325X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI325X (AMD CDNA 3) on DeepSeek V4 Pro 1.6T. 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 posts 370 tok/s/GPU for $1.06 per million tokens at 13 tok/s/user on DeepSeek V4 Pro 1.6T; MI325X posts 356 tok/s/GPU for $0.99. MI325X is 7% cheaper per token; H200 delivers 4% more tok/s/GPU.
Throughput at 20 tok/s/user on DeepSeek V4 Pro 1.6T: H200 hits 341 tok/s/GPU, MI325X hits 178. Per-million costs land at $1.15 and $2.04 respectively. H200 is 78% cheaper per token; H200 delivers 92% more tok/s/GPU.
H200 / MI325X on DeepSeek V4 Pro 1.6T at 27 tok/s/user: 305 / 95 tok/s/GPU, $1.29 / $3.76 per million tokens. H200 is 191% cheaper per token; H200 delivers 221% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:370.3MI325X:356.2 | H200:341.1MI325X:177.5 | H200:304.7MI325X:94.9 |
| Cost ($/M tok) | H200:$1.060MI325X:$0.992 | H200:$1.146MI325X:$2.042 | H200:$1.293MI325X:$3.764 |
| tok/s/MW | H200:270293MI325X:210776 | H200:248956MI325X:105035 | H200:222404MI325X:56183 |
| Concurrency | H200:~117MI325X:~113 | H200:~75MI325X:~37 | H200:~48MI325X:~14 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.