Kimi K2.5/K2.6 1T — B200 vs MI300X
Head-to-head AI inference benchmark comparison of B200 (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.
B200 posts 526 tok/s/GPU for $1.04 per million tokens at 40 tok/s/user on Kimi K2.5/K2.6 1T; MI300X posts 70 tok/s/GPU for $4.44. B200 is 328% cheaper per token; B200 delivers 647% more tok/s/GPU.
Throughput at 43 tok/s/user on Kimi K2.5/K2.6 1T: B200 hits 491 tok/s/GPU, MI300X hits 62. Per-million costs land at $1.11 and $5.09 respectively. B200 is 357% cheaper per token; B200 delivers 688% more tok/s/GPU.
B200 / MI300X on Kimi K2.5/K2.6 1T at 46 tok/s/user: 456 / 55 tok/s/GPU, $1.19 / $5.81 per million tokens. B200 is 387% cheaper per token; B200 delivers 735% more tok/s/GPU. (Numbers reflect the default 1k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:526.3MI300X:70.4 | B200:490.5MI300X:62.3 | B200:456.1MI300X:54.6 |
| Cost ($/M tok) | B200:$1.038MI300X:$4.441 | B200:$1.114MI300X:$5.089 | B200:$1.194MI300X:$5.813 |
| tok/s/MW | B200:242512MI300X:39352 | B200:226037MI300X:34789 | B200:210186MI300X:30525 |
| Concurrency | B200:~55MI300X:~7 | B200:~49MI300X:~6 | B200:~42MI300X:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.