diff --git a/src/chat/replyer/group_generator.py b/src/chat/replyer/group_generator.py
index 073338a2..2b43ddb0 100644
--- a/src/chat/replyer/group_generator.py
+++ b/src/chat/replyer/group_generator.py
@@ -1009,6 +1009,9 @@ class DefaultReplyer:
                 prompt
             )
 
+            # 移除 content 前后的换行符和空格
+            content = content.strip()
+
             logger.info(f"使用 {model_name} 生成回复内容: {content}")
         return content, reasoning_content, model_name, tool_calls
 
diff --git a/src/chat/replyer/private_generator.py b/src/chat/replyer/private_generator.py
index 747fe0c4..4c60be9d 100644
--- a/src/chat/replyer/private_generator.py
+++ b/src/chat/replyer/private_generator.py
@@ -930,6 +930,8 @@ class PrivateReplyer:
             content, (reasoning_content, model_name, tool_calls) = await self.express_model.generate_response_async(
                 prompt
             )
+            
+            content = content.strip()
 
             logger.info(f"使用 {model_name} 生成回复内容: {content}")
         return content, reasoning_content, model_name, tool_calls
diff --git a/src/plugins/built_in/memory/build_memory.py b/src/plugins/built_in/memory/build_memory.py
index 11dcbebc..3c1b4dc5 100644
--- a/src/plugins/built_in/memory/build_memory.py
+++ b/src/plugins/built_in/memory/build_memory.py
@@ -128,9 +128,9 @@ class GetMemoryTool(BaseTool):
         content_parts = [f"问题：{question}"]
         
         if memory_answer:
-            content_parts.append(f"记忆仓库答案：{memory_answer}")
+            content_parts.append(f"对问题'{question}'，你回忆的信息是：{memory_answer}")
         else:
-            content_parts.append(f"记忆仓库：对问题'{question}'，没有什么印象")
+            content_parts.append(f"对问题'{question}'，没有什么印象")
         
         if chat_answer:
             content_parts.append(f"对问题'{question}'，基于聊天记录的回答：{chat_answer}")
diff --git a/template/model_config_template.toml b/template/model_config_template.toml
index bc9313cc..830b27ca 100644
--- a/template/model_config_template.toml
+++ b/template/model_config_template.toml
@@ -35,9 +35,9 @@ name = "SiliconFlow"
 base_url = "https://api.siliconflow.cn/v1"
 api_key = "your-siliconflow-api-key"
 client_type = "openai"
-max_retry = 2
+max_retry = 3
 timeout = 120
-retry_interval = 10
+retry_interval = 5
 
 
 [[models]] # 模型（可以配置多个）
@@ -49,11 +49,22 @@ price_out = 8.0                    # 输出价格（用于API调用统计，单
 #force_stream_mode = true          # 强制流式输出模式（若模型不支持非流式输出，请取消该注释，启用强制流式输出，若无该字段，默认值为false）
 
 [[models]]
-model_identifier = "deepseek-ai/DeepSeek-V3"
-name = "siliconflow-deepseek-v3"
+model_identifier = "deepseek-ai/DeepSeek-V3.2-Exp"
+name = "siliconflow-deepseek-v3.2"
 api_provider = "SiliconFlow"
 price_in = 2.0
-price_out = 8.0
+price_out = 3.0
+[models.extra_params] # 可选的额外参数配置
+enable_thinking = false # 不启用思考
+
+[[models]]
+model_identifier = "deepseek-ai/DeepSeek-V3.2-Exp"
+name = "siliconflow-deepseek-v3.2-think"
+api_provider = "SiliconFlow"    
+price_in = 2.0
+price_out = 3.0
+[models.extra_params] # 可选的额外参数配置
+enable_thinking = true # 不启用思考
 
 [[models]]
 model_identifier = "deepseek-ai/DeepSeek-R1"
@@ -63,15 +74,6 @@ price_in = 4.0
 price_out = 16.0
 
 
-[[models]]
-model_identifier = "Qwen/Qwen3-8B"
-name = "qwen3-8b"
-api_provider = "SiliconFlow"
-price_in = 0
-price_out = 0
-[models.extra_params] # 可选的额外参数配置
-enable_thinking = false # 不启用思考
-
 [[models]]
 model_identifier = "Qwen/Qwen3-30B-A3B-Instruct-2507"
 name = "qwen3-30b"
@@ -80,8 +82,8 @@ price_in = 0.7
 price_out = 2.8
 
 [[models]]
-model_identifier = "Qwen/Qwen2.5-VL-72B-Instruct"
-name = "qwen2.5-vl-72b"
+model_identifier = "Qwen/Qwen3-VL-30B-A3B-Instruct"
+name = "qwen3-vl-30"
 api_provider = "SiliconFlow"
 price_in = 4.13
 price_out = 4.13
@@ -102,12 +104,12 @@ price_out = 0
 
 
 [model_task_config.utils] # 在麦麦的一些组件中使用的模型，例如表情包模块，取名模块，关系模块，麦麦的情绪变化等，是麦麦必须的模型
-model_list = ["siliconflow-deepseek-v3"] # 使用的模型列表，每个子项对应上面的模型名称(name)
+model_list = ["siliconflow-deepseek-v3.2"] # 使用的模型列表，每个子项对应上面的模型名称(name)
 temperature = 0.2                        # 模型温度，新V3建议0.1-0.3
 max_tokens = 2048                         # 最大输出token数
 
 [model_task_config.utils_small] # 在麦麦的一些组件中使用的小模型，消耗量较大，建议使用速度较快的小模型
-model_list = ["qwen3-8b","qwen3-30b"]
+model_list = ["qwen3-30b"]
 temperature = 0.7
 max_tokens = 2048
 
@@ -117,17 +119,17 @@ temperature = 0.7
 max_tokens = 800
 
 [model_task_config.replyer] # 首要回复模型，还用于表达器和表达方式学习
-model_list = ["siliconflow-deepseek-v3","siliconflow-deepseek-r1"]
+model_list = ["siliconflow-deepseek-v3.2-think","siliconflow-deepseek-r1","siliconflow-deepseek-v3.2"]
 temperature = 0.3                        # 模型温度，新V3建议0.1-0.3
 max_tokens = 800
 
 [model_task_config.planner] #决策：负责决定麦麦该什么时候回复的模型
-model_list = ["siliconflow-deepseek-v3"]
+model_list = ["siliconflow-deepseek-v3.2"]
 temperature = 0.3
 max_tokens = 800
 
 [model_task_config.vlm] # 图像识别模型
-model_list = ["qwen2.5-vl-72b"]
+model_list = ["qwen3-vl-30"]
 max_tokens = 256
 
 [model_task_config.voice] # 语音识别模型
@@ -140,16 +142,16 @@ model_list = ["bge-m3"]
 #------------LPMM知识库模型------------
 
 [model_task_config.lpmm_entity_extract] # 实体提取模型
-model_list = ["siliconflow-deepseek-v3"]
+model_list = ["siliconflow-deepseek-v3.2"]
 temperature = 0.2
 max_tokens = 800
 
 [model_task_config.lpmm_rdf_build] # RDF构建模型
-model_list = ["siliconflow-deepseek-v3"]
+model_list = ["siliconflow-deepseek-v3.2"]
 temperature = 0.2
 max_tokens = 800
 
 [model_task_config.lpmm_qa] # 问答模型
-model_list = ["qwen3-30b"]
+model_list = ["siliconflow-deepseek-v3.2"]
 temperature = 0.7
 max_tokens = 800