Kimi K2.5/K2.6/K2.7-Code 1T · GPU comparison

Kimi K2.5/K2.6/K2.7-Code 1T — GB300 NVL72 vs MI355X

Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) 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.

GB300 NVL72 posts 7015 tok/s/GPU for $0.11 per million tokens at 46 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T; MI355X posts 726 tok/s/GPU for $0.56. GB300 NVL72 is 431% cheaper per token; GB300 NVL72 delivers 866% more tok/s/GPU.

Throughput at 68 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: GB300 NVL72 hits 1181 tok/s/GPU, MI355X hits 554. Per-million costs land at $0.78 and $0.74 respectively. MI355X is 5% cheaper per token; GB300 NVL72 delivers 113% more tok/s/GPU.

GB300 NVL72 / MI355X on Kimi K2.5/K2.6/K2.7-Code 1T at 91 tok/s/user: 420 / 419 tok/s/GPU, $1.72 / $0.98 per million tokens. MI355X is 75% cheaper per token; throughput per GPU is essentially tied. (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.)

View performance-per-dollar view →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/gpu)
GB300 NVL72:7015.2MI355X:726.2
GB300 NVL72:1181.2MI355X:553.9
GB300 NVL72:419.6MI355X:418.8
Cost ($/M tok)
GB300 NVL72:$0.106MI355X:$0.565
GB300 NVL72:$0.776MI355X:$0.741
GB300 NVL72:$1.722MI355X:$0.982
tok/s/MW
GB300 NVL72:3340569MI355X:274039
GB300 NVL72:562471MI355X:209003
GB300 NVL72:199818MI355X:158047
Concurrency
GB300 NVL72:~1691MI355X:~35
GB300 NVL72:~312MI355X:~17
GB300 NVL72:~79MI355X:~9

Inference Performance

Inference performance metrics across different models, hardware configurations, and serving parameters.