Kimi K2.5/K2.6/K2.7-Code 1T — MI300X vs MI325X
Head-to-head AI inference benchmark comparison of MI300X (AMD CDNA 3) 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.
Near the low end of the 16–48 tok/s/user interactivity band, at 24 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: MI300X runs 117 tok/s/GPU at $2.63/M tokens, MI325X runs 144 at $2.50/M. MI325X is 5% cheaper per token; MI325X delivers 22% more tok/s/GPU.
Setting 32 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, MI300X produces 97 tok/s/GPU ($3.22 per million tokens) and MI325X produces 131 ($2.72). MI325X is 18% cheaper per token; MI325X delivers 35% more tok/s/GPU.
At 40 tok/s/user interactivity on Kimi K2.5/K2.6/K2.7-Code 1T, MI300X delivers 70 tok/s/GPU at $4.44 per million tokens; MI325X delivers 89 tok/s/GPU at $4.03. MI325X is 10% cheaper per token; MI325X delivers 26% more tok/s/GPU at this point. (Numbers reflect the default 1k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MI300X:117.3MI325X:143.6 | MI300X:97.2MI325X:131.1 | MI300X:70.4MI325X:88.7 |
| Cost ($/M tok) | MI300X:$2.627MI325X:$2.495 | MI300X:$3.224MI325X:$2.721 | MI300X:$4.441MI325X:$4.028 |
| tok/s/MW | MI300X:65505MI325X:65880 | MI300X:54323MI325X:60121 | MI300X:39352MI325X:40676 |
| Concurrency | MI300X:~20MI325X:~24 | MI300X:~12MI325X:~17 | MI300X:~7MI325X:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.