docs: add KTransformers CPU offloading inference guide

Add KTransformers as a recommended inference option for MiMo-V2-Flash.

KTransformers enables efficient deployment on consumer-grade hardware by offloading MoE expert computations to CPU while keeping other components on GPU. With 4× RTX 5090 + 2× AMD EPYC 9355, it achieves up to 35.7 tokens/s decode speed.

Benchmarks: https://ktransformers.net/benchmarks#MiMo-V2-Flash-FP8-TP4

README.md

@@ -250,6 +250,12 @@ curl -i http://localhost:9001/v1/chat/completions \
         }'
 ```
 
+### Inference with KTransformers (CPU Offloading)
+
+[KTransformers](https://github.com/kvcache-ai/ktransformers) enables efficient MiMo-V2-Flash deployment on consumer-grade hardware by offloading MoE expert computations to CPU, built on top of SGLang. With **4× RTX 5090 + 2× AMD EPYC 9355**, it achieves up to **35.7 tokens/s** decode speed.
+
+For quick start and benchmarks, visit [KTransformers](https://ktransformers.net/zh/benchmarks#MiMo-V2-Flash-FP8-TP4).
+
 ### Notifications
 
 #### 1. System prompt