diff --git a/docs/mddocs/Quickstart/llama_cpp_npu_portable_zip_quickstart.md b/docs/mddocs/Quickstart/llama_cpp_npu_portable_zip_quickstart.md
index 9212ec0f..da2aeb63 100644
--- a/docs/mddocs/Quickstart/llama_cpp_npu_portable_zip_quickstart.md
+++ b/docs/mddocs/Quickstart/llama_cpp_npu_portable_zip_quickstart.md
@@ -65,6 +65,10 @@ You could then use cli tool to run GGUF models on Intel NPU through running `lla
 llama-cli-npu.exe -m DeepSeek-R1-Distill-Qwen-7B-Q6_K.gguf -n 32 --prompt "What is AI?"
 ```
 
+> [!Note]
+> 
+> - The supported maximum number of input tokens is 960, and maximum sequence length for both input and output tokens is 1024 currently.
+
 ## Troubleshooting
 
 ### `L0 pfnCreate2 result: ZE_RESULT_ERROR_INVALID_ARGUMENT, code 0x78000004` error
diff --git a/docs/mddocs/Quickstart/llama_cpp_npu_portable_zip_quickstart.zh-CN.md b/docs/mddocs/Quickstart/llama_cpp_npu_portable_zip_quickstart.zh-CN.md
index df5a012b..8620bd43 100644
--- a/docs/mddocs/Quickstart/llama_cpp_npu_portable_zip_quickstart.zh-CN.md
+++ b/docs/mddocs/Quickstart/llama_cpp_npu_portable_zip_quickstart.zh-CN.md
@@ -65,6 +65,10 @@ IPEX-LLM 提供了 llama.cpp 的相关支持以在 Intel NPU 上运行 GGUF 模
 llama-cli-npu.exe -m DeepSeek-R1-Distill-Qwen-7B-Q6_K.gguf -n 32 --prompt "What is AI?"
 ```
 
+> [!Note]
+> 
+> - 目前支持的输入token数上限是960，输入和输出总token数上限是1024。
+
 ## 故障排除
 
 ### `L0 pfnCreate2 result: ZE_RESULT_ERROR_INVALID_ARGUMENT, code 0x78000004` 报错