diff --git a/dao/model.go b/dao/model.go
index 56bda81..8568f58 100644
--- a/dao/model.go
+++ b/dao/model.go
@@ -23,6 +23,14 @@ type FunctionModel struct {
 	//System   string `gorm:"column:system"`   //系统功能,对应模型系统参数
 }
 
+type ModelToken struct {
+	gorm.Model
+	ModelID   uint `gorm:"column:model_id"`   //模型id
+	UserID    uint `gorm:"column:user_id"`    //用户id
+	SessionID uint `gorm:"column:session_id"` //会话id
+	Token     uint `gorm:"column:token"`      //模型使用token
+}
+
 // 创建模型
 func CreateModel(userID uint, modelType, url, parameter, description string) (error, uint) {
 	model := Model{UserID: userID, Type: modelType, Url: url, Parameter: parameter, Description: description}
@@ -157,3 +165,34 @@ func FindFunctionModelByFunction(function string) []FunctionModel {
 	}
 	return models
 }
+
+func CreateModelToken(userID uint, modelID uint, sessionID, token uint) (error, uint) {
+	modelToken := ModelToken{UserID: userID, ModelID: modelID, Token: token, SessionID: sessionID}
+	var res *gorm.DB
+	if proto.Config.SERVER_SQL_LOG {
+		res = DB.Debug().Create(&modelToken)
+	} else {
+		res = DB.Create(&modelToken)
+	}
+	return res.Error, modelToken.ID
+}
+
+func FindModelTokenByModelID(modelID uint) []ModelToken {
+	var modelToken []ModelToken
+	if proto.Config.SERVER_SQL_LOG {
+		DB.Debug().Where("model_id = ?", modelID).Find(&modelToken)
+	} else {
+		DB.Where("model_id = ?", modelID).Find(&modelToken)
+	}
+	return modelToken
+}
+
+func FindModelTokenByUserID(userID uint) []ModelToken {
+	var modelToken []ModelToken
+	if proto.Config.SERVER_SQL_LOG {
+		DB.Debug().Where("user_id = ?", userID).Find(&modelToken)
+	} else {
+		DB.Where("user_id = ?", userID).Find(&modelToken)
+	}
+	return modelToken
+}
diff --git a/service/doubao.go b/service/doubao.go
index 4249e14..17679e1 100644
--- a/service/doubao.go
+++ b/service/doubao.go
@@ -108,6 +108,7 @@ func DouBaoV2(modelParam proto.ModelParam, imCtx *proto.IMParamContext) {
 	}(stream)
 
 	answer := ""
+	tokens := 0
 	for {
 		recv, err3 := stream.Recv()
 		if err3 == io.EOF {
@@ -123,6 +124,7 @@ func DouBaoV2(modelParam proto.ModelParam, imCtx *proto.IMParamContext) {
 		choices := recv.Choices[0]
 		if choices.FinishReason == proto.FinishReasonStop {
 			answer += choices.Delta.Content
+			tokens += recv.Usage.TotalTokens
 			break
 		} else {
 			answer += choices.Delta.Content
@@ -133,6 +135,12 @@ func DouBaoV2(modelParam proto.ModelParam, imCtx *proto.IMParamContext) {
 	if err != nil {
 		log.Println("spark create message error:", err)
 	}
+	//将本次请求的tokens存入数据库
+	err2, _ := dao.CreateModelToken(uint(imCtx.UserID), uint(imCtx.ModelID), uint(imCtx.SessionID), uint(tokens))
+	if err2 != nil {
+		log.Println("create model token error:", err2)
+	}
+
 }
 
 func doubaoToGeneralMassageAndSendMsgQueue(data *model.ChatCompletionStreamResponse, channel string, SessionID, UserID int) {
diff --git a/service/spark.go b/service/spark.go
index b634574..fa5fd5c 100644
--- a/service/spark.go
+++ b/service/spark.go
@@ -211,6 +211,7 @@ func SparkV2(modelParam proto.ModelParam, imCtx *proto.IMParamContext) {
 	}()
 
 	answer := ""
+	tokens := 0
 
 	//接收消息
 	for {
@@ -233,6 +234,7 @@ func SparkV2(modelParam proto.ModelParam, imCtx *proto.IMParamContext) {
 		if data.Header.Status == 2 {
 			//最后一个结果
 			answer += data.Payload.Choices.Text[0].Content
+			tokens += data.Payload.Usage.Text.TotalTokens
 		} else {
 			answer += data.Payload.Choices.Text[0].Content
 		}
@@ -248,6 +250,14 @@ func SparkV2(modelParam proto.ModelParam, imCtx *proto.IMParamContext) {
 	} else {
 		log.Println("spark create message success,message id:", msgID)
 	}
+	//存入tokens消耗信息
+	err3, modelTokenID := dao.CreateModelToken(uint(imCtx.UserID), uint(imCtx.ModelID), uint(imCtx.SessionID), uint(tokens))
+	if err3 != nil {
+		log.Println("spark create message error:", err3)
+	} else {
+		log.Println("spark create model tokens id:", modelTokenID)
+	}
+
 }
 
 func SparkToGeneralMassageAndSendMsgQueue(data *SparkResponse, channel string, SessionID int, UserID int) {