Kimi K2.5/K2.6/K2.7-Code 1T — H200 vs MI355X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI355X (AMD CDNA 4) 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.
Throughput at 46 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: H200 hits 350 tok/s/GPU, MI355X hits 428. Per-million costs land at $1.12 and $0.96 respectively. MI355X is 16% cheaper per token; MI355X delivers 22% more tok/s/GPU.
H200 / MI355X on Kimi K2.5/K2.6/K2.7-Code 1T at 59 tok/s/user: 273 / 298 tok/s/GPU, $1.42 / $1.37 per million tokens. MI355X is 4% cheaper per token; MI355X delivers 9% more tok/s/GPU.
Toward the upper edge of the 34–85 tok/s/user interactivity band, at 72 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: H200 runs 217 tok/s/GPU at $1.79/M tokens, MI355X runs 222 at $1.88/M. H200 is 5% cheaper per token; MI355X delivers 2% more tok/s/GPU. (Numbers reflect the default 1k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:350.3MI355X:427.6 | H200:272.7MI355X:298.1 | H200:216.6MI355X:221.7 |
| Cost ($/M tok) | H200:$1.116MI355X:$0.961 | H200:$1.421MI355X:$1.371 | H200:$1.795MI355X:$1.879 |
| tok/s/MW | H200:202502MI355X:161361 | H200:157639MI355X:112503 | H200:125187MI355X:83662 |
| Concurrency | H200:~31MI355X:~19 | H200:~19MI355X:~10 | H200:~12MI355X:~6 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.