Kimi K2.5/K2.6/K2.7-Code 1T — H200 vs MI325X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI325X (AMD CDNA 3) on Kimi K2.5/K2.6/K2.7-Code 1T. 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 38 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, H200 produces 451 tok/s/GPU ($0.89 per million tokens) and MI325X produces 100 ($3.58). H200 is 302% cheaper per token; H200 delivers 350% more tok/s/GPU.
At 42 tok/s/user interactivity on Kimi K2.5/K2.6/K2.7-Code 1T, H200 delivers 396 tok/s/GPU at $1.01 per million tokens; MI325X delivers 79 tok/s/GPU at $4.49. H200 is 346% cheaper per token; H200 delivers 400% more tok/s/GPU at this point.
H200 posts 341 tok/s/GPU for $1.14 per million tokens at 47 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T; MI325X posts 62 tok/s/GPU for $5.81. H200 is 409% cheaper per token; H200 delivers 452% more tok/s/GPU. (Numbers reflect the default 1k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:450.5MI325X:100.1 | H200:396.1MI325X:79.2 | H200:340.8MI325X:61.8 |
| Cost ($/M tok) | H200:$0.891MI325X:$3.581 | H200:$1.006MI325X:$4.490 | H200:$1.142MI325X:$5.811 |
| tok/s/MW | H200:260427MI325X:45923 | H200:228951MI325X:36311 | H200:196996MI325X:28337 |
| Concurrency | H200:~50MI325X:~11 | H200:~40MI325X:~8 | H200:~30MI325X:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.