diff --git a/packages/model-runtime/src/core/streams/qwen.test.ts b/packages/model-runtime/src/core/streams/qwen.test.ts
index 85d98d09f3..923282cf66 100644
--- a/packages/model-runtime/src/core/streams/qwen.test.ts
+++ b/packages/model-runtime/src/core/streams/qwen.test.ts
@@ -479,6 +479,146 @@ describe('QwenAIStream', () => {
       `data: [{"function":{"arguments":"","name":"get_weather"},"id":"call_123","index":0,"type":"function"}]\n\n`,
     ]);
   });
+
+  it('should handle mixed text content followed by streaming tool calls (DeepSeek style)', async () => {
+    // This test simulates the stream pattern from DeepSeek models via Qwen API
+    // where text content is streamed first, followed by incremental tool call chunks
+    const mockOpenAIStream = new ReadableStream({
+      start(controller) {
+        // Text content chunks with role in first chunk
+        controller.enqueue({
+          choices: [
+            {
+              delta: { content: '看来', role: 'assistant' },
+              finish_reason: null,
+              index: 0,
+            },
+          ],
+          id: 'chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075',
+          model: 'deepseek-v3',
+          object: 'chat.completion.chunk',
+          created: 1767574524,
+        });
+        controller.enqueue({
+          choices: [
+            {
+              delta: { content: '我的' },
+              finish_reason: null,
+              index: 0,
+            },
+          ],
+          id: 'chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075',
+          model: 'deepseek-v3',
+          object: 'chat.completion.chunk',
+          created: 1767574524,
+        });
+        controller.enqueue({
+          choices: [
+            {
+              delta: { content: '函数调用格式有误。' },
+              finish_reason: null,
+              index: 0,
+            },
+          ],
+          id: 'chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075',
+          model: 'deepseek-v3',
+          object: 'chat.completion.chunk',
+          created: 1767574524,
+        });
+
+        // First tool call chunk with id, name, and partial arguments
+        controller.enqueue({
+          choices: [
+            {
+              delta: {
+                tool_calls: [
+                  {
+                    id: 'call_ff00c42325d74b979990cb',
+                    type: 'function',
+                    function: {
+                      name: 'modelscope-time____get_current_time____mcp',
+                      arguments: '{"',
+                    },
+                    index: 0,
+                  },
+                ],
+              },
+              finish_reason: null,
+              index: 0,
+            },
+          ],
+          id: 'chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075',
+          model: 'deepseek-v3',
+          object: 'chat.completion.chunk',
+          created: 1767574524,
+        });
+
+        // Subsequent tool call chunk with only incremental arguments (no id)
+        controller.enqueue({
+          choices: [
+            {
+              delta: {
+                tool_calls: [
+                  {
+                    type: 'function',
+                    function: {
+                      arguments: 'timezone":"America/New_York"}',
+                    },
+                    index: 0,
+                  },
+                ],
+              },
+              finish_reason: null,
+              index: 0,
+            },
+          ],
+          id: 'chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075',
+          model: 'deepseek-v3',
+          object: 'chat.completion.chunk',
+          created: 1767574524,
+        });
+
+        controller.close();
+      },
+    });
+
+    const onTextMock = vi.fn();
+    const onToolCallMock = vi.fn();
+
+    const protocolStream = QwenAIStream(mockOpenAIStream, {
+      callbacks: {
+        onText: onTextMock,
+        onToolsCalling: onToolCallMock,
+      },
+    });
+
+    const decoder = new TextDecoder();
+    const chunks = [];
+
+    // @ts-ignore
+    for await (const chunk of protocolStream) {
+      chunks.push(decoder.decode(chunk, { stream: true }));
+    }
+
+    // Verify complete chunks array
+    expect(chunks).toEqual([
+      'id: chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075\n',
+      'event: text\n',
+      'data: "看来"\n\n',
+      'id: chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075\n',
+      'event: text\n',
+      'data: "我的"\n\n',
+      'id: chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075\n',
+      'event: text\n',
+      'data: "函数调用格式有误。"\n\n',
+      'id: chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075\n',
+      'event: tool_calls\n',
+      'data: [{"function":{"arguments":"{\\"","name":"modelscope-time____get_current_time____mcp"},"id":"call_ff00c42325d74b979990cb","index":0,"type":"function"}]\n\n',
+      'id: chatcmpl-4f901cb2-91bc-9763-a2c8-3ed58e9f4075\n',
+      'event: tool_calls\n',
+      'data: [{"function":{"arguments":"timezone\\":\\"America/New_York\\"}","name":null},"id":"call_ff00c42325d74b979990cb","index":0,"type":"function"}]\n\n',
+    ]);
+  });
 });
 
 describe('transformQwenStream', () => {
diff --git a/packages/model-runtime/src/core/streams/qwen.ts b/packages/model-runtime/src/core/streams/qwen.ts
index 62c892d0d5..733e5f4543 100644
--- a/packages/model-runtime/src/core/streams/qwen.ts
+++ b/packages/model-runtime/src/core/streams/qwen.ts
@@ -70,8 +70,18 @@ export const transformQwenStream = (
 
   if (item.delta?.tool_calls) {
     return {
-      data: item.delta.tool_calls.map(
-        (value, index): StreamToolCallChunkData => ({
+      data: item.delta.tool_calls.map((value, index): StreamToolCallChunkData => {
+        // Store first tool call's info in streamContext for subsequent chunks
+        // (similar pattern to OpenAI stream handling)
+        if (streamContext && !streamContext.tool && value.id && value.function?.name) {
+          streamContext.tool = {
+            id: value.id,
+            index: typeof value.index !== 'undefined' ? value.index : index,
+            name: value.function.name,
+          };
+        }
+
+        return {
           // Qwen models may send tool_calls in two separate chunks:
           // 1. First chunk: {id, name} without arguments
           // 2. Second chunk: {id, arguments} without name
@@ -81,11 +91,13 @@ export const transformQwenStream = (
             arguments: value.function?.arguments ?? '',
             name: value.function?.name ?? null,
           },
-          id: value.id || generateToolCallId(index, value.function?.name),
+          // For incremental chunks without id, use the stored tool id from streamContext
+          id:
+            value.id || streamContext?.tool?.id || generateToolCallId(index, value.function?.name),
           index: typeof value.index !== 'undefined' ? value.index : index,
           type: value.type || 'function',
-        }),
-      ),
+        };
+      }),
       id: chunk.id,
       type: 'tool_calls',
     } as StreamProtocolToolCallChunk;
diff --git a/src/app/[variants]/(main)/chat/features/Conversation/Header/index.tsx b/src/app/[variants]/(main)/chat/features/Conversation/Header/index.tsx
index de874baa3b..13c17a4679 100644
--- a/src/app/[variants]/(main)/chat/features/Conversation/Header/index.tsx
+++ b/src/app/[variants]/(main)/chat/features/Conversation/Header/index.tsx
@@ -1,5 +1,6 @@
 'use client';
 
+import { isDesktop } from '@lobechat/const';
 import { Flexbox } from '@lobehub/ui';
 import { cssVar } from 'antd-style';
 import { memo } from 'react';
@@ -22,7 +23,7 @@ const Header = memo(() => {
       }
       right={
         <Flexbox align={'center'} horizontal style={{ backgroundColor: cssVar.colorBgContainer }}>
-          <WorkingDirectory />
+          {isDesktop && <WorkingDirectory />}
           <NotebookButton />
           <ShareButton />
           <HeaderActions />