diff --git a/src/agentscope/model/_dashscope_model.py b/src/agentscope/model/_dashscope_model.py
index c17a08e703..99e2e66512 100644
--- a/src/agentscope/model/_dashscope_model.py
+++ b/src/agentscope/model/_dashscope_model.py
@@ -368,6 +368,10 @@ async def _parse_dashscope_stream_response(
                     input_tokens=chunk.usage.input_tokens,
                     output_tokens=chunk.usage.output_tokens,
                     time=(datetime.now() - start_datetime).total_seconds(),
+                    prompt_tokens_details=chunk.usage.get(
+                        "prompt_tokens_details",
+                        None,
+                    ),
                 )
 
             parsed_chunk = ChatResponse(
@@ -469,6 +473,10 @@ async def _parse_dashscope_generation_response(
                 input_tokens=response.usage.input_tokens,
                 output_tokens=response.usage.output_tokens,
                 time=(datetime.now() - start_datetime).total_seconds(),
+                prompt_tokens_details=response.usage.get(
+                    "prompt_tokens_details",
+                    None,
+                ),
             )
 
         parsed_response = ChatResponse(
diff --git a/src/agentscope/model/_model_usage.py b/src/agentscope/model/_model_usage.py
index fcc76120ff..5e314be6d3 100644
--- a/src/agentscope/model/_model_usage.py
+++ b/src/agentscope/model/_model_usage.py
@@ -19,5 +19,8 @@ class ChatUsage(DictMixin):
     time: float
     """The time used in seconds."""
 
+    prompt_tokens_details: dict = field(default_factory=dict)
+    """The details of cached tokens. """
+
     type: Literal["chat"] = field(default_factory=lambda: "chat")
     """The type of the usage, must be `chat`."""
diff --git a/src/agentscope/tracing/_attributes.py b/src/agentscope/tracing/_attributes.py
index 991481b5af..5a92a5411b 100644
--- a/src/agentscope/tracing/_attributes.py
+++ b/src/agentscope/tracing/_attributes.py
@@ -68,6 +68,9 @@ class SpanAttributes:
     GEN_AI_USAGE_OUTPUT_TOKENS = GenAIAttributes.GEN_AI_USAGE_OUTPUT_TOKENS
     """The gen ai usage output tokens."""
 
+    GEN_AI_USAGE_CACHED_TOKENS = "gen_ai.dash.usage.cached_tokens"
+    """The gen ai usage cached tokens."""
+
     # GenAI Message Attributes
     GEN_AI_INPUT_MESSAGES = GenAIAttributes.GEN_AI_INPUT_MESSAGES
     """The gen ai input messages."""
diff --git a/src/agentscope/tracing/_extractor.py b/src/agentscope/tracing/_extractor.py
index 6ca71ea87e..5af90a562f 100644
--- a/src/agentscope/tracing/_extractor.py
+++ b/src/agentscope/tracing/_extractor.py
@@ -378,6 +378,13 @@ def _get_llm_response_attributes(
         attributes[
             SpanAttributes.GEN_AI_USAGE_OUTPUT_TOKENS
         ] = chat_response.usage.output_tokens
+        if (
+            hasattr(chat_response.usage, "prompt_tokens_details")
+            and chat_response.usage.prompt_tokens_details
+        ):
+            attributes[
+                SpanAttributes.GEN_AI_USAGE_CACHED_TOKENS
+            ] = _serialize_to_str(chat_response.usage.prompt_tokens_details)
 
     output_messages = _get_llm_output_messages(chat_response)
     if output_messages: