Kimi K2.5/K2.6 1T — GB200 NVL72 vs MI300X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI300X (AMD CDNA 3) on Kimi K2.5/K2.6 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.
At 60 tok/s/user on Kimi K2.5/K2.6 1T, GB200 NVL72 delivers 1855 tok/s/GPU at $0.33 per million tokens; MI300X hasn't been benchmarked at this target.
GB200 NVL72 hits 584 tok/s/GPU for $1.18 per million tokens at 103 tok/s/user on Kimi K2.5/K2.6 1T. No MI300X data at this operating point.
GB200 NVL72: 133 tok/s/GPU, $4.69 per million tokens at 146 tok/s/user on Kimi K2.5/K2.6 1T. MI300X is unmeasured here. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:1854.8MI300X:— | GB200 NVL72:584.3MI300X:— | GB200 NVL72:133.2MI300X:— |
| Cost ($/M tok) | GB200 NVL72:$0.328MI300X:— | GB200 NVL72:$1.180MI300X:— | GB200 NVL72:$4.687MI300X:— |
| tok/s/MW | GB200 NVL72:883236MI300X:— | GB200 NVL72:278250MI300X:— | GB200 NVL72:63416MI300X:— |
| Concurrency | GB200 NVL72:~667MI300X:— | GB200 NVL72:~116MI300X:— | GB200 NVL72:~13MI300X:— |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.